Carbon vs Silicon Consumption — 2025/2026 comparison

Executive summary

This is the short version: human consumption is still dominated by shelter, health, food, finance, and services; the modeled silicon basket is dominated by compute, HBM, networking, storage, and advanced packaging.

Human total PCE (2025)

$20.95T

+5.3% vs. 2024

Official U.S. personal consumption expenditures.

Human goods vs. services

31.1% / 68.9%

$6.51T goods · $14.45T services

Human consumption is service-heavy.

Silicon modeled basket

$298.0B

2025 modeled AI-server-related basket

Modeled from TrendForce's AI-server-related industry value anchor.

Human top bucket

Housing and utilities

$3.80T

Shelter dominates the current human basket.

Silicon top group

Accelerators & compute

$123.67B

Compute dominates the modeled silicon basket.

Scale ratio

70.3×

Human total PCE / silicon basket

Not an apples-to-apples accounting ratio, but a useful scale check.

Five takeaways

Current official U.S. human consumption is still overwhelmingly about shelter, health care, everyday goods, finance, and services: total 2025 PCE was $20.95T, with services at 68.9% of the total. H3
The clean human top-level partition is led by Housing and utilities ($3.80T) and Health care ($3.55T), not by flashy discretionary gadgets. H3
The silicon-side proxy is much smaller in total dollars but much more concentrated: the model anchors to a $298.0B 2025 AI-server-related basket, and Accelerators & compute alone takes 41.5% of it. S1 S2
The fastest-moving silicon line is not generic compute alone but memory attached to compute: TrendForce says 2025 AI-chip shipments would boost HBM demand by more than 130%, which is why HBM is the second-biggest modeled basket group. S2
The tweet's intuition mostly works as a capital-allocation metaphor: human society spends mostly on shelter/health/food/services, while the current AI stack spends mostly on compute/memory/bandwidth. But the comparison is between end-use household consumption and a modeled industrial procurement basket, so it should be read as a resource-allocation analogy, not a common-accounting statement. H2 H3 S1 S3

What the tweet gets right

The capital market really has redirected an enormous amount of spending toward the hardware stack that keeps models training and serving: accelerators, HBM, networking, and packaging. In that sense, “silicon-based consumption” is a useful metaphor. S1 S2 S3 S5

But the metaphor breaks if you read it literally. Human consumption is still orders of magnitude larger, much more service-heavy, and much less concentrated. AI infrastructure is an industrial bottleneck basket, not a household basket. H3 S1 S3

Put differently: humans mostly buy shelter, health, calories, mobility, finance, and social functioning. Silicon mostly “buys” throughput, bandwidth, memory, and power density. H3 S2 S3

Methodology and caveats

The whole page is only useful if the boundaries are explicit. This section makes them explicit.

Human side: what is official, and what is not

The official human side comes from BEA/FRED Table 2.4.5 for calendar year 2025. BEA's own FAQ says Table 2.4.5U is the most detailed current time-series PCE table, while the benchmark-year PCE Bridge is where deeper commodity composition lives. H1 H2 H3

That creates a practical limitation: BEA does not currently publish 100 separate current goods-only leaf lines. So the page uses two human views:

Clean official partition: 16 non-overlapping top-level branches from the current table.
Full 100-series published view: exactly 100 current published rows after excluding 13 high-level structural totals and nonprofit subtraction lines.

The 100-series published view is very useful for ranking current official lines by size, but it mixes leaf lines and roll-ups, so those 100 rows should not be summed. H2 H3

Silicon side: why it is modeled

There is no official statistical agency publishing a “silicon lifeform consumption basket.” So the page constructs a proxy from the strongest observable 2025 bottleneck basket: AI-server-related industry value. TrendForce put that value at US$298B for 2025; IDC separately reported the worldwide 2025 server market at US$444.1B, up 80.4% year over year. S1 S3

IDC's taxonomy is important because it defines what a server system contains: processors, motherboard, memory, internal disk/flash, bundled OS, power supplies, and network interfaces. The model then expands from that base using source-led signals around HBM, networking, packaging, and storage. S3 S5 S6 S7

The silicon 100-row table is therefore a modeled allocation, not a published revenue ranking. Amounts sum exactly to US$298B by construction; trends are directional source-led estimates, not audited line-item revenues. S1 S2 S3

Assumptions used in the build

Human side: use the current official BEA/FRED annual table for 2025 (Table 2.4.5) for all published values and 2024 comparisons.
Human side: because BEA does not publish 100 separate current goods-only leaf series, the page shows two views: a clean official top-level partition plus a 100-series current published table that mixes leaf lines and roll-ups.
Human side: the 100-series current published table excludes lines 1, 2, 3, 4, 8, 13, 19, 25, 47, 48, 111, 112, and 113 to leave exactly 100 current published series.
Silicon side: allocate the 2025 TrendForce AI-server-related industry value anchor (US$298B) across 100 procurement categories.
Silicon side: use IDC's server-system taxonomy to define the component universe, then bias weights using the strongest 2025 signals from TrendForce, IDC networking, SIA, and SEMI.
Silicon trends are source-led directional estimates for each modeled bucket; they are not audited segment revenues.

Human / carbon-based consumption

Official current U.S. consumption is still dominated by housing, health care, broad services, and food. Goods matter, but services are almost 69% of total 2025 PCE. H3

Clean official partition: top human buckets

The 16-bucket partition is the best non-overlapping current view for macro comparison. H3

Housing and utilities

$3.80T

Health care

$3.55T

Other services

$1.78T

Other nondurable goods

$1.73T

Financial services and insurance

$1.69T

Food and beverages purchased for off-premises consumption

$1.53T

Food services and accommodations

$1.50T

Recreation services

$819.2B

Motor vehicles and parts

$758.8B

Transportation services

$703.7B

Largest current goods-only leaf lines

This is the goods-only view that most closely matches the user's original wording. It is smaller and less detailed than the full 100-series current table. H3

Food and nonalcoholic beverages purchased for off-premises consumption

$1,295.1B

Pharmaceutical and other medical products

$729.4B

Video, audio, photographic, and information processing equipment and media

$427.6B

New motor vehicles

$408.5B

Motor vehicle fuels, lubricants, and fluids

$393.1B

Recreational items

$334.7B

Furniture and furnishings

$293.9B

Women's and girls' clothing

$255.9B

Alcoholic beverages purchased for off-premises consumption

$233.9B

Net purchases of used motor vehicles

$219.1B

Personal care products

$213.6B

Household supplies

$198.9B

Men's and boys' clothing

$153.7B

Motor vehicle parts and accessories

$131.2B

Magazines, newspapers, and stationery

$127.0B

Interpretation

The human basket is not mostly gadgets or even goods. It is mostly shelter, medical care, food, financial intermediation, and other services. The high-dollar lines are recurring, necessary, and broadly distributed across the population. H3

Even the biggest goods lines are really “maintenance of life and mobility” lines: groceries, pharmaceuticals, vehicles, fuels, household supplies, and basic apparel. Discretionary goods exist, but they sit below the necessity stack in total dollars. H3

Trend pattern

Human top-level buckets mostly move in the mid-single digits, with outliers driven by commodity prices or insurance/medical pricing. In 2025, total PCE rose +5.3%, goods rose +3.9%, and services rose +6.0%. H3

That is very different from the silicon basket, where memory, networking, and compute bottlenecks can post high-double-digit or triple-digit growth in a single year. S2 S5

Open the clean 16-bucket human partition table

Human consumption: clean 16-bucket partition (non-overlapping)

This is the cleanest current official partition available in Table 2.4.5. These 16 first-level branches do not overlap and do sum to total PCE.

Rank ↕	Published line ↕	Human top-level bucket ↕	2025 spend ↕	2024 spend ↕	Trend vs. 2024 ↕	Share of total PCE ↕	Reasoning ↕	Sources ↕
1	BEA line 49	Housing and utilities	$3,801.2B	$3,603.6B	+5.5%	18.14%	Shelter is the core human consumption function, and it absorbs the largest recurring share of household budgets.	H3
2	BEA line 60	Health care	$3,551.6B	$3,315.0B	+7.1%	16.95%	Medical spending is supported by demographics, chronic care demand, and high labor and facility intensity.	H3
3	BEA line 95	Other services	$1,777.9B	$1,675.8B	+6.1%	8.48%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
4	BEA line 39	Other nondurable goods	$1,728.6B	$1,633.1B	+5.8%	8.25%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
5	BEA line 86	Financial services and insurance	$1,689.8B	$1,560.4B	+8.3%	8.06%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
6	BEA line 26	Food and beverages purchased for off-premises consumption	$1,529.9B	$1,479.6B	+3.4%	7.30%	Groceries are universal, repetitive spending, so even low single-digit growth compounds into very large dollar totals.	H3
7	BEA line 81	Food services and accommodations	$1,495.2B	$1,429.0B	+4.6%	7.14%	Eating away from home layers wages, rent, and service margins onto basic calorie demand.	H3
8	BEA line 76	Recreation services	$819.2B	$777.0B	+5.4%	3.91%	Households continue to spend on leisure, media, and entertainment when income and confidence hold up.	H3
9	BEA line 4	Motor vehicles and parts	$758.8B	$732.8B	+3.6%	3.62%	Transportation needs keep autos and parts near the top of the durable-goods stack.	H3
10	BEA line 68	Transportation services	$703.7B	$660.5B	+6.5%	3.36%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
11	BEA line 13	Recreational goods and vehicles	$697.6B	$663.7B	+5.1%	3.33%	Entertainment hardware and sports gear remain meaningful discretionary goods categories.	H3
12	BEA line 111	Final consumption expenditures of nonprofit institutions serving households (NPISHs)	$608.2B	$613.1B	-0.8%	2.90%	Nonprofit-provided household services are economically real consumption even when paid indirectly.	H3
13	BEA line 30	Clothing and footwear	$561.8B	$530.2B	+6.0%	2.68%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
14	BEA line 8	Furnishings and durable household equipment	$507.3B	$490.0B	+3.5%	2.42%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
15	BEA line 36	Gasoline and other energy goods	$422.4B	$440.5B	-4.1%	2.02%	Fuel spending moves with miles driven and commodity prices, making it one of the more volatile household goods lines.	H3
16	BEA line 19	Other durable goods	$301.6B	$291.6B	+3.5%	1.44%	Small but durable household and personal items add up because of broad usage.	H3

Open the official goods-only highlight table (top 20 current leaf lines)

Human goods-only highlights: top 20 current leaf categories

Because the current BEA time-series table does not publish 100 separate goods-only leaves, this table shows the largest current official goods-only leaf categories.

BEA line ↕	Goods-only leaf category ↕	Family ↕	2025 spend ↕	2024 spend ↕	Trend vs. 2024 ↕	Share of goods total ↕	Reasoning ↕
BEA line 27	Food and nonalcoholic beverages purchased for off-premises consumption	Food at home	$1,295.1B	$1,250.1B	+3.6%	19.90%	Groceries are universal, repetitive spending, so even low single-digit growth compounds into very large dollar totals.
BEA line 40	Pharmaceutical and other medical products	Other nondurables	$729.4B	$677.7B	+7.6%	11.21%	Prescriptions and medical products combine steady utilization with high branded-product mix.
BEA line 14	Video, audio, photographic, and information processing equipment and media	Recreation durables	$427.6B	$402.7B	+6.2%	6.57%	Electronics stay large because computing, media, and device-refresh spending are now everyday household behaviors.
BEA line 5	New motor vehicles	Vehicles & parts	$408.5B	$394.9B	+3.5%	6.28%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.
BEA line 37	Motor vehicle fuels, lubricants, and fluids	Energy goods	$393.1B	$414.3B	-5.1%	6.04%	Fuel spending moves with miles driven and commodity prices, making it one of the more volatile household goods lines.
BEA line 41	Recreational items	Other nondurables	$334.7B	$313.7B	+6.7%	5.14%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.
BEA line 9	Furniture and furnishings	Home durables	$293.9B	$281.9B	+4.3%	4.52%	Home setup and replacement cycles keep this tied to household formation and renovation activity.
BEA line 32	Women's and girls' clothing	Apparel & footwear	$255.9B	$240.7B	+6.3%	3.93%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.
BEA line 28	Alcoholic beverages purchased for off-premises consumption	Food at home	$233.9B	$228.7B	+2.3%	3.59%	Staple consumption happens every week, making grocery categories some of the most dependable spending lines.
BEA line 6	Net purchases of used motor vehicles	Vehicles & parts	$219.1B	$211.5B	+3.6%	3.37%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.
BEA line 43	Personal care products	Other nondurables	$213.6B	$203.4B	+5.0%	3.28%	Personal care is frequent, habitual spending with broad category penetration.
BEA line 42	Household supplies	Other nondurables	$198.9B	$191.1B	+4.1%	3.06%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.
BEA line 33	Men's and boys' clothing	Apparel & footwear	$153.7B	$144.1B	+6.6%	2.36%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.
BEA line 7	Motor vehicle parts and accessories	Vehicles & parts	$131.2B	$126.4B	+3.8%	2.02%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.
BEA line 45	Magazines, newspapers, and stationery	Other nondurables	$127.0B	$119.0B	+6.8%	1.95%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.
BEA line 15	Sporting equipment, supplies, guns, and ammunition	Recreation durables	$126.9B	$121.8B	+4.2%	1.95%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.
BEA line 35	Other clothing materials and footwear	Apparel & footwear	$124.6B	$119.3B	+4.4%	1.91%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.
BEA line 44	Tobacco	Other nondurables	$111.6B	$115.2B	-3.2%	1.71%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.
BEA line 20	Jewelry and watches	Other durables	$106.0B	$104.6B	+1.4%	1.63%	Small but durable household and personal items add up because of broad usage.
BEA line 16	Sports and recreational vehicles	Recreation durables	$100.9B	$99.4B	+1.5%	1.55%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.

Open the full 100 current published human series table

Human consumption: full top-100 current published series

To get to 100 current published series, the page removes 13 high-level structural totals/subtractions and keeps the remaining published rows. This table mixes 75 leaf lines and 25 roll-ups, so it should not be summed.

Rank ↕	Published line ↕	Published category ↕	Type ↕	Family ↕	2025 spend ↕	2024 spend ↕	Trend vs. 2024 ↕	Share of total PCE ↕	Reasoning ↕	Sources ↕
1	BEA line 49	Housing and utilities	Roll-up	Housing & utilities	$3,801.2B	$3,603.6B	+5.5%	18.14%	Shelter is the core human consumption function, and it absorbs the largest recurring share of household budgets.	H3
2	BEA line 60	Health care	Roll-up	Health care services	$3,551.6B	$3,315.0B	+7.1%	16.95%	Medical spending is supported by demographics, chronic care demand, and high labor and facility intensity.	H3
3	BEA line 50	Housing	Roll-up	Housing & utilities	$3,311.0B	$3,147.0B	+5.2%	15.80%	Shelter is the core human consumption function, and it absorbs the largest recurring share of household budgets.	H3
4	BEA line 52	Imputed rental of owner-occupied nonfarm housing	Leaf	Housing & utilities	$2,500.3B	$2,374.2B	+5.3%	11.93%	Owner-occupied housing is counted as a service households consume, so shelter dominates even without a cash rent check.	H3
5	BEA line 65	Hospital and nursing home services	Roll-up	Health care services	$1,869.2B	$1,742.3B	+7.3%	8.92%	High-acuity care is expensive, labor-intensive, and used steadily across an aging population.	H3
6	BEA line 95	Other services	Roll-up	Other services	$1,777.9B	$1,675.8B	+6.1%	8.48%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
7	BEA line 39	Other nondurable goods	Roll-up	Other nondurables	$1,728.6B	$1,633.1B	+5.8%	8.25%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
8	BEA line 86	Financial services and insurance	Roll-up	Financial services & insurance	$1,689.8B	$1,560.4B	+8.3%	8.06%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
9	BEA line 61	Outpatient services	Roll-up	Health care services	$1,682.4B	$1,572.7B	+7.0%	8.03%	Routine care, specialty visits, diagnostics, and therapy make this a high-frequency medical spend.	H3
10	BEA line 66	Hospitals	Leaf	Health care services	$1,600.2B	$1,493.6B	+7.1%	7.64%	High-acuity care is expensive, labor-intensive, and used steadily across an aging population.	H3
11	BEA line 26	Food and beverages purchased for off-premises consumption	Roll-up	Food at home	$1,529.9B	$1,479.6B	+3.4%	7.30%	Groceries are universal, repetitive spending, so even low single-digit growth compounds into very large dollar totals.	H3
12	BEA line 81	Food services and accommodations	Roll-up	Food away from home	$1,495.2B	$1,429.0B	+4.6%	7.14%	Eating away from home layers wages, rent, and service margins onto basic calorie demand.	H3
13	BEA line 82	Food services	Roll-up	Food away from home	$1,297.3B	$1,230.8B	+5.4%	6.19%	Eating away from home layers wages, rent, and service margins onto basic calorie demand.	H3
14	BEA line 27	Food and nonalcoholic beverages purchased for off-premises consumption	Leaf	Food at home	$1,295.1B	$1,250.1B	+3.6%	6.18%	Groceries are universal, repetitive spending, so even low single-digit growth compounds into very large dollar totals.	H3
15	BEA line 83	Purchased meals and beverages	Leaf	Food away from home	$1,260.0B	$1,195.9B	+5.4%	6.01%	Eating away from home layers wages, rent, and service margins onto basic calorie demand.	H3
16	BEA line 87	Financial services	Roll-up	Financial services & insurance	$1,147.5B	$1,041.5B	+10.2%	5.48%	Payments, lending, asset management, and deposit intermediation scale with balances and transaction volumes.	H3
17	BEA line 62	Physician services	Leaf	Health care services	$836.7B	$793.3B	+5.5%	3.99%	Routine care, specialty visits, diagnostics, and therapy make this a high-frequency medical spend.	H3
18	BEA line 76	Recreation services	Roll-up	Recreation services	$819.2B	$777.0B	+5.4%	3.91%	Households continue to spend on leisure, media, and entertainment when income and confidence hold up.	H3
19	BEA line 51	Rental of tenant-occupied nonfarm housing	Leaf	Housing & utilities	$776.8B	$740.3B	+4.9%	3.71%	Shelter is the most persistent household bill; rents reset slowly but on a very large base.	H3
20	BEA line 40	Pharmaceutical and other medical products	Leaf	Other nondurables	$729.4B	$677.7B	+7.6%	3.48%	Prescriptions and medical products combine steady utilization with high branded-product mix.	H3
21	BEA line 68	Transportation services	Roll-up	Transportation services	$703.7B	$660.5B	+6.5%	3.36%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
22	BEA line 64	Paramedical services	Leaf	Health care services	$647.4B	$593.7B	+9.1%	3.09%	Routine care, specialty visits, diagnostics, and therapy make this a high-frequency medical spend.	H3
23	BEA line 88	Financial services furnished without payment	Leaf	Financial services & insurance	$614.5B	$565.5B	+8.7%	2.93%	Payments, lending, asset management, and deposit intermediation scale with balances and transaction volumes.	H3
24	BEA line 30	Clothing and footwear	Roll-up	Apparel & footwear	$561.8B	$530.2B	+6.0%	2.68%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
25	BEA line 90	Insurance	Roll-up	Financial services & insurance	$542.4B	$518.9B	+4.5%	2.59%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
26	BEA line 89	Financial service charges, fees, and commissions	Leaf	Financial services & insurance	$533.0B	$476.0B	+12.0%	2.54%	Payments, lending, asset management, and deposit intermediation scale with balances and transaction volumes.	H3
27	BEA line 55	Household utilities	Roll-up	Housing & utilities	$490.3B	$456.7B	+7.4%	2.34%	Shelter and utility services are unavoidable household expenses, so they stay large even when growth is modest.	H3
28	BEA line 31	Garments	Roll-up	Apparel & footwear	$437.2B	$410.9B	+6.4%	2.09%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
29	BEA line 14	Video, audio, photographic, and information processing equipment and media	Leaf	Recreation durables	$427.6B	$402.7B	+6.2%	2.04%	Electronics stay large because computing, media, and device-refresh spending are now everyday household behaviors.	H3
30	BEA line 36	Gasoline and other energy goods	Roll-up	Energy goods	$422.4B	$440.5B	-4.1%	2.02%	Fuel spending moves with miles driven and commodity prices, making it one of the more volatile household goods lines.	H3
31	BEA line 5	New motor vehicles	Leaf	Vehicles & parts	$408.5B	$394.9B	+3.5%	1.95%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.	H3
32	BEA line 69	Motor vehicle services	Roll-up	Transportation services	$395.2B	$368.6B	+7.2%	1.89%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
33	BEA line 37	Motor vehicle fuels, lubricants, and fluids	Leaf	Energy goods	$393.1B	$414.3B	-5.1%	1.88%	Fuel spending moves with miles driven and commodity prices, making it one of the more volatile household goods lines.	H3
34	BEA line 100	Education services	Roll-up	Other services	$363.1B	$352.2B	+3.1%	1.73%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
35	BEA line 106	Social services and religious activities	Leaf	Other services	$355.4B	$334.7B	+6.2%	1.70%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
36	BEA line 57	Electricity and gas	Roll-up	Housing & utilities	$343.4B	$317.0B	+8.3%	1.64%	Utilities are recurring necessities, with weather and regulated pricing driving year-to-year movement.	H3
37	BEA line 41	Recreational items	Leaf	Other nondurables	$334.7B	$313.7B	+6.7%	1.60%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
38	BEA line 96	Communication	Roll-up	Other services	$319.3B	$309.1B	+3.3%	1.52%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
39	BEA line 72	Public transportation	Roll-up	Transportation services	$308.5B	$291.9B	+5.7%	1.47%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
40	BEA line 77	Membership clubs, sports centers, parks, theaters, and museums	Leaf	Recreation services	$306.0B	$291.2B	+5.1%	1.46%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
41	BEA line 104	Professional and other services	Leaf	Other services	$297.7B	$277.9B	+7.1%	1.42%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
42	BEA line 9	Furniture and furnishings	Leaf	Home durables	$293.9B	$281.9B	+4.3%	1.40%	Home setup and replacement cycles keep this tied to household formation and renovation activity.	H3
43	BEA line 109	Foreign travel by U.S. residents	Leaf	Other services	$287.3B	$269.2B	+6.7%	1.37%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
44	BEA line 93	Net health insurance	Leaf	Financial services & insurance	$283.5B	$270.1B	+4.9%	1.35%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
45	BEA line 70	Motor vehicle maintenance and repair	Leaf	Transportation services	$276.7B	$257.5B	+7.5%	1.32%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
46	BEA line 67	Nursing homes	Leaf	Health care services	$269.0B	$248.7B	+8.2%	1.28%	High-acuity care is expensive, labor-intensive, and used steadily across an aging population.	H3
47	BEA line 58	Electricity	Leaf	Housing & utilities	$267.5B	$253.4B	+5.6%	1.28%	Utilities are recurring necessities, with weather and regulated pricing driving year-to-year movement.	H3
48	BEA line 32	Women's and girls' clothing	Leaf	Apparel & footwear	$255.9B	$240.7B	+6.3%	1.22%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
49	BEA line 105	Personal care and clothing services	Leaf	Other services	$241.7B	$233.1B	+3.7%	1.15%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
50	BEA line 28	Alcoholic beverages purchased for off-premises consumption	Leaf	Food at home	$233.9B	$228.7B	+2.3%	1.12%	Staple consumption happens every week, making grocery categories some of the most dependable spending lines.	H3
51	BEA line 79	Gambling	Leaf	Recreation services	$229.3B	$206.9B	+10.8%	1.09%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
52	BEA line 6	Net purchases of used motor vehicles	Leaf	Vehicles & parts	$219.1B	$211.5B	+3.6%	1.05%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.	H3
53	BEA line 43	Personal care products	Leaf	Other nondurables	$213.6B	$203.4B	+5.0%	1.02%	Personal care is frequent, habitual spending with broad category penetration.	H3
54	BEA line 110	Less: Expenditures in the United States by nonresidents	Leaf	Other services	$208.9B	$216.6B	-3.5%	1.00%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
55	BEA line 101	Higher education	Leaf	Other services	$208.2B	$206.3B	+0.9%	0.99%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
56	BEA line 42	Household supplies	Leaf	Other nondurables	$198.9B	$191.1B	+4.1%	0.95%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
57	BEA line 63	Dental services	Leaf	Health care services	$198.2B	$185.7B	+6.7%	0.95%	Routine care, specialty visits, diagnostics, and therapy make this a high-frequency medical spend.	H3
58	BEA line 85	Accommodations	Leaf	Food away from home	$197.9B	$198.2B	-0.2%	0.94%	Eating away from home layers wages, rent, and service margins onto basic calorie demand.	H3
59	BEA line 74	Air transportation	Leaf	Transportation services	$195.5B	$189.1B	+3.4%	0.93%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
60	BEA line 97	Telecommunication services	Leaf	Other services	$185.4B	$181.4B	+2.2%	0.88%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
61	BEA line 78	Audio-video, photographic, and information processing equipment services	Leaf	Recreation services	$172.4B	$171.7B	+0.4%	0.82%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
62	BEA line 33	Men's and boys' clothing	Leaf	Apparel & footwear	$153.7B	$144.1B	+6.6%	0.73%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
63	BEA line 56	Water supply and sanitation	Leaf	Housing & utilities	$146.8B	$139.6B	+5.2%	0.70%	Utilities are recurring necessities, with weather and regulated pricing driving year-to-year movement.	H3
64	BEA line 7	Motor vehicle parts and accessories	Leaf	Vehicles & parts	$131.2B	$126.4B	+3.8%	0.63%	Vehicles are large-ticket replacement purchases, so price and financing conditions matter as much as unit demand.	H3
65	BEA line 91	Life insurance	Leaf	Financial services & insurance	$130.3B	$127.6B	+2.2%	0.62%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
66	BEA line 45	Magazines, newspapers, and stationery	Leaf	Other nondurables	$127.0B	$119.0B	+6.8%	0.61%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
67	BEA line 15	Sporting equipment, supplies, guns, and ammunition	Leaf	Recreation durables	$126.9B	$121.8B	+4.2%	0.61%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
68	BEA line 35	Other clothing materials and footwear	Leaf	Apparel & footwear	$124.6B	$119.3B	+4.4%	0.59%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
69	BEA line 107	Household maintenance	Leaf	Other services	$122.3B	$116.1B	+5.3%	0.58%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
70	BEA line 71	Other motor vehicle services	Leaf	Transportation services	$118.5B	$111.1B	+6.6%	0.57%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
71	BEA line 99	Internet access	Leaf	Other services	$116.9B	$110.6B	+5.7%	0.56%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
72	BEA line 44	Tobacco	Leaf	Other nondurables	$111.6B	$115.2B	-3.2%	0.53%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
73	BEA line 80	Other recreational services	Leaf	Recreation services	$111.5B	$107.2B	+4.0%	0.53%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
74	BEA line 94	Net motor vehicle and other transportation insurance	Leaf	Financial services & insurance	$108.8B	$102.9B	+5.7%	0.52%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
75	BEA line 20	Jewelry and watches	Leaf	Other durables	$106.0B	$104.6B	+1.4%	0.51%	Small but durable household and personal items add up because of broad usage.	H3
76	BEA line 73	Ground transportation	Leaf	Transportation services	$105.6B	$96.0B	+10.1%	0.50%	Transport demand follows commuting, travel, freight-linked ticketing, and maintenance of personal mobility.	H3
77	BEA line 16	Sports and recreational vehicles	Leaf	Recreation durables	$100.9B	$99.4B	+1.5%	0.48%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
78	BEA line 21	Therapeutic appliances and equipment	Leaf	Other durables	$99.4B	$95.6B	+4.0%	0.47%	Prescriptions and medical products combine steady utilization with high branded-product mix.	H3
79	BEA line 10	Household appliances	Leaf	Home durables	$89.7B	$88.1B	+1.8%	0.43%	Home setup and replacement cycles keep this tied to household formation and renovation activity.	H3
80	BEA line 103	Commercial and vocational schools	Leaf	Other services	$88.1B	$82.8B	+6.3%	0.42%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
81	BEA line 108	Net foreign travel	Roll-up	Other services	$78.4B	$52.6B	+48.9%	0.37%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
82	BEA line 59	Natural gas	Leaf	Housing & utilities	$75.9B	$63.6B	+19.4%	0.36%	Utilities are recurring necessities, with weather and regulated pricing driving year-to-year movement.	H3
83	BEA line 102	Nursery, elementary, and secondary schools	Leaf	Other services	$66.8B	$63.0B	+6.0%	0.32%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
84	BEA line 11	Glassware, tableware, and household utensils	Leaf	Home durables	$62.0B	$59.1B	+4.9%	0.30%	Home setup and replacement cycles keep this tied to household formation and renovation activity.	H3
85	BEA line 12	Tools and equipment for house and garden	Leaf	Home durables	$61.8B	$61.0B	+1.4%	0.30%	Home setup and replacement cycles keep this tied to household formation and renovation activity.	H3
86	BEA line 23	Luggage and similar personal items	Leaf	Other durables	$51.1B	$48.0B	+6.5%	0.24%	Small but durable household and personal items add up because of broad usage.	H3
87	BEA line 84	Food furnished to employees (including military)	Leaf	Food away from home	$37.3B	$34.9B	+6.9%	0.18%	Restaurant and lodging spend is recurring and labor-heavy, so dollar totals stay high.	H3
88	BEA line 24	Telephone and facsimile equipment	Leaf	Other durables	$36.1B	$34.8B	+3.7%	0.17%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
89	BEA line 17	Recreational books	Leaf	Recreation durables	$33.1B	$31.1B	+6.5%	0.16%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
90	BEA line 53	Rental value of farm dwellings	Leaf	Housing & utilities	$30.5B	$29.2B	+4.7%	0.15%	Shelter and utility services are unavoidable household expenses, so they stay large even when growth is modest.	H3
91	BEA line 38	Fuel oil and other fuels	Leaf	Energy goods	$29.3B	$26.3B	+11.4%	0.14%	Fuel spending moves with miles driven and commodity prices, making it one of the more volatile household goods lines.	H3
92	BEA line 34	Children's and infants' clothing	Leaf	Apparel & footwear	$27.6B	$26.1B	+5.9%	0.13%	Apparel is a steady refresh category shaped by seasons, fashion cycles, and value-channel pricing.	H3
93	BEA line 92	Net household insurance	Leaf	Financial services & insurance	$19.8B	$18.3B	+8.1%	0.09%	Insurance spending rises with underlying asset values, claims costs, and medical utilization.	H3
94	BEA line 98	Postal and delivery services	Leaf	Other services	$17.1B	$17.1B	-0.2%	0.08%	Communications, education, professional help, and social services are diffuse but collectively very large.	H3
95	BEA line 46	Net expenditures abroad by U.S. residents	Leaf	Other nondurables	$13.4B	$13.1B	+2.8%	0.06%	Fast-turn, everyday-use items create large aggregate spend despite modest ticket sizes.	H3
96	BEA line 18	Musical instruments	Leaf	Recreation durables	$9.1B	$8.7B	+4.0%	0.04%	Experience and hobby spending remains durable when consumers feel confident and keep discretionary budgets open.	H3
97	BEA line 22	Educational books	Leaf	Other durables	$8.9B	$8.5B	+5.0%	0.04%	Small but durable household and personal items add up because of broad usage.	H3
98	BEA line 75	Water transportation	Leaf	Transportation services	$7.4B	$6.8B	+8.1%	0.04%	Utilities are recurring necessities, with weather and regulated pricing driving year-to-year movement.	H3
99	BEA line 54	Group housing	Leaf	Housing & utilities	$3.4B	$3.3B	+3.3%	0.02%	Shelter and utility services are unavoidable household expenses, so they stay large even when growth is modest.	H3
100	BEA line 29	Food produced and consumed on farms	Leaf	Food at home	$0.9B	$0.8B	+10.6%	0.00%	Staple consumption happens every week, making grocery categories some of the most dependable spending lines.	H3

Silicon / AI-cluster consumption

The modeled silicon basket asks a different question: if you treat today's AI stack as a resource-consuming system, where do the dollars actually go? The short answer is compute first, then HBM, then networking/interconnect, then the supporting stack around them. S1 S2 S3 S5

Modeled non-overlapping group summary

The 9-group structure is the clean apples-to-apples macro view of the silicon basket. S1 S2 S3 S5 S7

Accelerators & compute

$123.67B

HBM

$49.17B

Networking & interconnect

$26.82B

Host CPUs & support silicon

$22.35B

Storage

$17.88B

Standard memory

$17.88B

Advanced packaging, substrates & test

$16.39B

Optics & photonics

$13.41B

Power, cooling, mechanicals & integration

$10.43B

Top modeled silicon categories

The largest modeled line items are not generic “chips” in the abstract. They are specific spend buckets around hyperscaler accelerators, HBM attach, and high-speed interconnect. This is a concentrated, bottleneck-driven basket. S2 S5 S13

Hyperscaler training accelerators

$21.28B

Accelerators & compute

Hyperscaler inference accelerators

$17.26B

Accelerators & compute

OEM training accelerators

$12.65B

Accelerators & compute

OEM inference accelerators

$10.93B

Accelerators & compute

HBM3e high-capacity stacks

$10.43B

HBM

HBM3e mainstream stacks

$8.94B

HBM

Sovereign-cloud accelerators

$8.63B

Accelerators & compute

HBM3e for custom ASICs

$7.70B

HBM

Custom CSP training ASICs

$7.48B

Accelerators & compute

Custom CSP inference ASICs

$6.33B

Accelerators & compute

China-compliant accelerators

$5.52B

Accelerators & compute

HBM for sovereign-cloud systems

$5.46B

HBM

Full-rack accelerator blades

$5.18B

Accelerators & compute

Modular accelerator trays

$4.83B

Accelerators & compute

800G AI NICs

$4.56B

Networking & interconnect

Reasoning behind the basket

Compute dominates because the highest-value AI systems still clear around the most powerful accelerators. TrendForce's 2025/2026 work also implies a step-up in full-rack systems and a continued surge in custom ASIC investment. S1 S2

HBM is second because advanced AI chips are memory-bandwidth hungry, and TrendForce says 2025 AI-chip shipments alone would boost HBM demand by more than 130%. Packaging and test matter because those memory-attached multi-die packages are hard to build and qualify. S2 S7

Networking and optics matter because larger model graphs mean larger clusters, and larger clusters mean more pressure on bandwidth, losslessness, latency, and optical reach. IDC's networking data and Broadcom's 2026 product set make that visible. S5 S13

Trend pattern

The silicon basket is much more explosive than the human basket. TrendForce said 2025 AI-server revenue was expected to rise by almost 48%; IDC showed the 2025 server market up 80.4%; datacenter Ethernet switch revenue hit US$32.5B for 2025; and HBM demand was projected to surge more than 130% from 2025 AI-chip shipments. S2 S3 S5

That kind of dispersion is why the page marks confidence levels on the modeled 100-row table. The big buckets are easier to anchor than the long tail. S1 S2 S3

Open the silicon 9-group summary table

Silicon basket: 9 modeled non-overlapping groups

These groups are designed to be non-overlapping and sum exactly to the US$298B AI-server-related basket anchor.

Silicon basket group ↕	Modeled 2025 amount ↕	Share of modeled basket ↕	Weighted trend ↕	Line items ↕
Accelerators & compute	$123.67B	41.50%	+54.5%	20
HBM	$49.17B	16.50%	+131.7%	10
Networking & interconnect	$26.82B	9.00%	+55.6%	15
Host CPUs & support silicon	$22.35B	7.50%	+31.7%	10
Storage	$17.88B	6.00%	+28.7%	10
Standard memory	$17.88B	6.00%	+42.8%	10
Advanced packaging, substrates & test	$16.39B	5.50%	+39.2%	10
Optics & photonics	$13.41B	4.50%	+69.0%	10
Power, cooling, mechanicals & integration	$10.43B	3.50%	+37.9%	5

Open the full 100 modeled silicon categories table

Silicon basket: full 100 modeled categories

These are modeled procurement buckets, not audited published segment revenues. Amounts are allocated from the US$298B basket anchor using source-led group weights and subgroup adjustments.

Rank ↕	Group ↕	Modeled silicon category ↕	Modeled 2025 amount ↕	Share of basket ↕	Trend vs. 2024 (modeled) ↕	Confidence ↕	Reasoning ↕	Sources ↕
1	Accelerators & compute	Hyperscaler training accelerators	$21.28B	7.14%	+58.0%	High	Largest single spend bucket: frontier-model training still clears price through the most expensive accelerator modules.	S1 S2 S3 S4 S12
2	Accelerators & compute	Hyperscaler inference accelerators	$17.26B	5.79%	+62.0%	High	Inference is scaling rapidly as deployed models serve more users, raising demand for dense accelerator fleets.	S1 S2 S3 S4 S12
3	Accelerators & compute	OEM training accelerators	$12.65B	4.25%	+44.0%	High	Enterprise and OEM training clusters trail hyperscaler scale but still buy high-ASP accelerator systems.	S1 S2 S3 S4 S12
4	Accelerators & compute	OEM inference accelerators	$10.93B	3.67%	+47.0%	High	Commercial deployments are broadening from pilots to production inference, lifting OEM accelerator demand.	S1 S2 S3 S4 S12
5	HBM	HBM3e high-capacity stacks	$10.43B	3.50%	+145.0%	High	The highest-bandwidth memory stacks attach to the most expensive training accelerators and capture the richest pricing.	S1 S2 S6 S7 S12
6	HBM	HBM3e mainstream stacks	$8.94B	3.00%	+130.0%	High	Mainstream HBM3e is scaling with the broad ramp of current-generation AI accelerators.	S1 S2 S6 S7 S12
7	Accelerators & compute	Sovereign-cloud accelerators	$8.63B	2.90%	+52.0%	High	National and regional AI buildouts are creating a second wave of high-end accelerator purchases.	S1 S2 S3 S4 S12
8	HBM	HBM3e for custom ASICs	$7.70B	2.58%	+150.0%	High	Custom AI ASICs are adopting HBM aggressively to keep model throughput high.	S1 S2 S6 S7 S12
9	Accelerators & compute	Custom CSP training ASICs	$7.48B	2.51%	+72.0%	High	Large cloud providers are increasing in-house silicon to control cost, power, and model-specific optimization.	S1 S2 S3 S4 S12
10	Accelerators & compute	Custom CSP inference ASICs	$6.33B	2.12%	+78.0%	Medium	Custom inference silicon grows faster where token economics reward workload-specific designs.	S1 S2 S3 S4 S12
11	Accelerators & compute	China-compliant accelerators	$5.52B	1.85%	+36.0%	Medium	Restrictions cap performance and complicate mix, but compliant products still command meaningful spend.	S1 S2 S3 S4 S12
12	HBM	HBM for sovereign-cloud systems	$5.46B	1.83%	+135.0%	High	Sovereign-cloud deployments increasingly require top-tier memory density, not just compute availability.	S1 S2 S6 S7 S12
13	Accelerators & compute	Full-rack accelerator blades	$5.18B	1.74%	+80.0%	Medium	Integrated rack-scale systems raise ticket size by bundling compute, networking, and power at the rack level.	S1 S2 S3 S4 S12
14	Accelerators & compute	Modular accelerator trays	$4.83B	1.62%	+51.0%	Medium	Modular sleds remain a common procurement unit for AI clusters outside the newest full-rack designs.	S1 S2 S3 S4 S12
15	Networking & interconnect	800G AI NICs	$4.56B	1.53%	+86.0%	High	Higher-speed NICs become mandatory as cluster size and model parallelism rise.	S3 S5 S13
16	HBM	HBM for China-compliant systems	$4.47B	1.50%	+80.0%	High	Export restrictions constrain the highest-end attach rates in some China-facing systems.	S1 S2 S6 S7 S12
17	Accelerators & compute	OAM accelerator modules	$4.37B	1.47%	+40.0%	Medium	OAM remains important for scale-up systems, though spend is being re-tilted toward integrated rack systems.	S1 S2 S3 S4 S12
18	Host CPUs & support silicon	x86 host CPUs for AI training nodes	$4.36B	1.46%	+32.0%	Medium	Accelerated servers still need powerful hosts for orchestration, preprocessing, and system balance.	S3 S4 S12
19	Advanced packaging, substrates & test	CoWoS / 2.5D packaging services	$4.26B	1.43%	+35.0%	Medium	Advanced package capacity remains a chokepoint for high-end AI silicon.	S2 S7 S12
20	Standard memory	High-capacity DDR5 RDIMMs	$3.84B	1.29%	+50.0%	Medium	CPU-side memory remains essential for data preparation, orchestration, and mixed workloads around the accelerator cluster.	S2 S6 S12
21	Accelerators & compute	SXM accelerator modules	$3.80B	1.27%	+38.0%	Medium	Legacy high-end module formats still matter, but roadmap mix is shifting toward newer rack-level designs.	S1 S2 S3 S4 S12
22	Networking & interconnect	400G AI NICs	$3.73B	1.25%	+46.0%	High	400G still ships heavily but is being displaced at the frontier by 800G designs.	S3 S5 S13
23	Storage	High-performance NVMe SSDs for training	$3.58B	1.20%	+30.0%	Medium	Training clusters need very fast local and shared storage for data staging and checkpointing.	S8 S9 S10 S11
24	HBM	HBM4 early sampling and pilots	$3.48B	1.17%	+200.0%	High	Early HBM4 work is small in absolute dollars but growing explosively as next-gen platforms prepare.	S1 S2 S6 S7 S12
25	Host CPUs & support silicon	x86 host CPUs for AI inference nodes	$3.24B	1.09%	+38.0%	Medium	Inference servers often use more CPU-side orchestration than pure training boxes.	S3 S4 S12
26	HBM	Capacity-reserved HBM wafer starts	$3.23B	1.08%	+125.0%	Medium	Reservation and allocation behavior remains important in a tight market, but growth is slower than delivered stacks.	S1 S2 S6 S7 S12
27	Accelerators & compute	PCIe accelerator cards	$3.22B	1.08%	+26.0%	Medium	PCIe add-in cards stay relevant for midrange deployment, but they are losing share to denser AI-server formats.	S1 S2 S3 S4 S12
28	Power, cooling, mechanicals & integration	High-density power supplies	$3.13B	1.05%	+30.0%	Medium	Power density climbs with accelerator racks, lifting value per node even without explosive unit growth.	S2 S3 S13
29	Optics & photonics	800G optical transceivers	$3.08B	1.04%	+62.0%	Medium	The shift to denser AI fabrics lifts transceiver value directly as links scale up.	S5 S13
30	Storage	Capacity NVMe SSDs for inference	$3.04B	1.02%	+26.0%	Medium	Inference fleets increasingly favor dense flash for fast retrieval and low-latency serving.	S8 S9 S10 S11
31	Standard memory	Standard DDR5 RDIMMs	$2.95B	0.99%	+42.0%	Medium	Standard DIMMs are still widely consumed but are less supply-constrained than HBM.	S2 S6 S12
32	Advanced packaging, substrates & test	3D-IC / 3.5D packaging services	$2.79B	0.94%	+60.0%	Medium	More vertically integrated package formats are growing from a smaller base.	S2 S7 S12
33	Networking & interconnect	SmartNICs and DPUs	$2.76B	0.93%	+52.0%	High	Infrastructure offload remains valuable for storage, security, and datapath efficiency.	S3 S5 S13
34	Accelerators & compute	Edge inference accelerators	$2.76B	0.93%	+30.0%	Medium	Edge AI spend is growing, but it is much smaller than datacenter-scale procurement.	S1 S2 S3 S4 S12
35	HBM	Pre-packaged HBM inventory buffers	$2.73B	0.92%	+100.0%	Low	Inventory buffers exist because lead times are strategic, though spend here grows slower than end demand.	S1 S2 S6 S7 S12
36	Host CPUs & support silicon	Arm host CPUs for AI nodes	$2.68B	0.90%	+44.0%	Medium	Arm server CPU adoption is increasing in cloud environments where power efficiency and platform control matter.	S3 S4 S12
37	Networking & interconnect	102.4T switch ASICs	$2.63B	0.88%	+90.0%	High	Very large switch silicon captures outsized value in the newest scale-out AI fabrics.	S3 S5 S13
38	Accelerators & compute	Scale-up fabric ASICs	$2.53B	0.85%	+73.0%	Medium	As clusters grow, scale-up fabrics become a direct spend line rather than a design afterthought.	S1 S2 S3 S4 S12
39	Storage	QLC SSDs for vector / retrieval stores	$2.50B	0.84%	+50.0%	Medium	Retrieval-augmented systems and AI data services make cheaper high-capacity flash more attractive.	S8 S9 S10 S11
40	Optics & photonics	1.6T optical transceivers	$2.41B	0.81%	+107.0%	Medium	1.6T is still early but is one of the fastest-ramping premium optical categories.	S5 S13
41	Power, cooling, mechanicals & integration	Voltage regulation and power shelves	$2.40B	0.81%	+35.0%	Medium	Rack-scale AI systems need much heavier in-rack power management than prior server generations.	S2 S3 S13
42	Host CPUs & support silicon	General server CPUs for control planes	$2.35B	0.79%	+20.0%	Medium	Control-plane workloads grow, but they are smaller and lower-ASP than main host CPU sockets.	S3 S4 S12
43	Standard memory	LPDDR / GDDR for midrange AI systems	$2.23B	0.75%	+35.0%	Medium	Midrange and specialized systems use graphics-style memory where cost or form factor matters.	S2 S6 S12
44	Advanced packaging, substrates & test	ABF substrates	$2.21B	0.74%	+30.0%	Medium	Premium substrates remain a necessary enabler for complex accelerator packages.	S2 S7 S12
45	Host CPUs & support silicon	CXL controller silicon	$2.12B	0.71%	+50.0%	Medium	CXL is still early but becoming more relevant as memory pooling and expansion move from concept to deployment.	S3 S4 S12
46	Networking & interconnect	51.2T switch ASICs	$2.07B	0.70%	+36.0%	Medium	51.2T remains a major installed base, but leading-edge demand is moving upward.	S3 S5 S13
47	Accelerators & compute	Accelerator bridge silicon	$2.07B	0.69%	+60.0%	Medium	Bridge silicon expands with multi-die and multi-accelerator packaging complexity.	S1 S2 S3 S4 S12
48	Power, cooling, mechanicals & integration	Liquid cooling cold plates / manifolds	$1.98B	0.67%	+50.0%	Medium	Liquid cooling adoption rises as rack power density breaks air-cooling limits.	S2 S3 S13
49	Standard memory	CXL memory expansion modules	$1.97B	0.66%	+63.0%	Medium	Disaggregated memory is still early, but it is one of the faster-growing memory-adjacent categories.	S2 S6 S12
50	Networking & interconnect	Low-latency fabric controllers	$1.94B	0.65%	+50.0%	Medium	AI networking prioritizes low jitter and lossless performance, supporting premium fabric silicon.	S3 S5 S13
51	Optics & photonics	Optical DSPs	$1.88B	0.63%	+60.0%	Medium	Higher-speed optics increasingly rely on advanced DSP content.	S5 S13
52	Storage	Boot and local scratch SSDs	$1.88B	0.63%	+18.0%	Medium	Every system needs boot and scratch storage, but this is a lower-ASP part of the stack.	S8 S9 S10 S11
53	Storage	Nearline HDDs for AI data lakes	$1.79B	0.60%	+22.0%	Medium	HDDs still dominate the cold and warm layers of large AI data repositories.	S8 S9 S10 S11
54	Host CPUs & support silicon	Chipsets and I/O hubs	$1.79B	0.60%	+22.0%	Medium	Support logic scales with server volume but does not enjoy the same pricing power as accelerators or memory.	S3 S4 S12
55	HBM	Legacy HBM2e/HBM3 support	$1.74B	0.58%	+40.0%	Low	Older HBM generations still ship but are being displaced quickly by newer, denser stacks.	S1 S2 S6 S7 S12
56	Accelerators & compute	Compute chiplet sets	$1.73B	0.58%	+66.0%	Medium	Chipletization improves yield and design modularity, making it a larger share of advanced compute value.	S1 S2 S3 S4 S12
57	Host CPUs & support silicon	Baseboard management controllers	$1.68B	0.56%	+24.0%	Medium	Every server still needs management and telemetry, but these controllers are a modest share of system value.	S3 S4 S12
58	Networking & interconnect	Ethernet retimers	$1.66B	0.56%	+42.0%	Medium	Signal integrity spending rises with lane speed, though retimers remain a supporting category.	S3 S5 S13
59	Advanced packaging, substrates & test	Silicon interposers	$1.64B	0.55%	+40.0%	Medium	Interposer demand rises with HBM attach and multi-die packaging.	S2 S7 S12
60	Standard memory	Pooled memory appliances	$1.61B	0.54%	+59.0%	Medium	Shared-memory designs are emerging where large inference state and data locality matter.	S2 S6 S12
61	Power, cooling, mechanicals & integration	In-rack CDUs and pump loops	$1.56B	0.53%	+54.0%	Medium	CDUs and pump loops grow quickly once operators adopt liquid cooling at cluster scale.	S2 S3 S13
62	Networking & interconnect	PCIe Gen6 switches	$1.52B	0.51%	+48.0%	Medium	PCIe fabric expansion matters more as accelerator and storage topologies grow denser.	S3 S5 S13
63	Storage	SSD controllers and storage silicon	$1.52B	0.51%	+20.0%	Medium	Controller spend follows SSD volumes and enterprise mix shifts.	S8 S9 S10 S11
64	Optics & photonics	Co-packaged optics modules	$1.48B	0.50%	+84.0%	Medium	CPO moves from demonstration to meaningful spend where power and reach constraints tighten.	S5 S13
65	Advanced packaging, substrates & test	Package test and burn-in	$1.48B	0.50%	+55.0%	Medium	AI parts demand much heavier test intensity because of performance, yield, and thermal requirements.	S2 S7 S12
66	Host CPUs & support silicon	Boot and service processors	$1.45B	0.49%	+17.0%	Low	These processors are operationally necessary but economically small.	S3 S4 S12
67	Standard memory	DIMM PMICs and buffers	$1.43B	0.48%	+35.0%	Medium	Module-side power and control silicon rises with server memory density but remains a support category.	S2 S6 S12
68	Accelerators & compute	I/O chiplet sets	$1.38B	0.46%	+56.0%	Low	I/O chiplets benefit from the same multi-die trend but at a lower dollar intensity than compute tiles.	S1 S2 S3 S4 S12
69	Power, cooling, mechanicals & integration	Chassis, racks, and final integration	$1.36B	0.46%	+25.0%	Medium	Mechanical integration grows with system complexity, but it still trails silicon and memory in value capture.	S2 S3 S13
70	Host CPUs & support silicon	Memory-expansion controllers	$1.34B	0.45%	+40.0%	Low	More complex memory topologies make specialized control logic more valuable.	S3 S4 S12
71	Host CPUs & support silicon	Replacement and buffer CPU inventory	$1.34B	0.45%	+13.0%	Low	Buffers are prudent in a tight market, but they are not the core driver of silicon consumption.	S3 S4 S12
72	Standard memory	Low-latency memory for inference caches	$1.25B	0.42%	+55.0%	Medium	Inference-serving architectures increasingly optimize around cache hit rates and low-latency retrieval.	S2 S6 S12
73	Storage	DPU-attached storage modules	$1.25B	0.42%	+35.0%	Medium	Data-path acceleration is increasingly pulled closer to storage in AI-oriented architectures.	S8 S9 S10 S11
74	Networking & interconnect	PCIe Gen6 retimers	$1.24B	0.42%	+44.0%	Medium	Retimers benefit from the same speed transition but at smaller absolute value.	S3 S5 S13
75	Optics & photonics	VCSEL / EML laser components	$1.21B	0.40%	+50.0%	Medium	Laser component demand follows module volumes and speed transitions.	S5 S13
76	Advanced packaging, substrates & test	Wafer probe and characterization	$1.15B	0.39%	+40.0%	Medium	Probe intensity rises with die complexity and qualification cycles.	S2 S7 S12
77	Networking & interconnect	Scale-out network interface silicon	$1.11B	0.37%	+55.0%	Medium	Interface silicon is needed everywhere, but premium value concentrates in NICs and switches.	S3 S5 S13
78	Storage	AI-native storage appliances	$1.07B	0.36%	+40.0%	Low	AI-specific storage systems are growing from a small base as vendors bundle performance around retrieval workloads.	S8 S9 S10 S11
79	Optics & photonics	Photodiodes and receiver arrays	$1.07B	0.36%	+44.0%	Medium	Receiver-side optics scale with transceiver and photonics deployment.	S5 S13
80	Standard memory	Server-memory PCB and module materials	$1.07B	0.36%	+30.0%	Low	Materials benefit from module volume, but pricing is less explosive than primary memory silicon.	S2 S6 S12
81	Networking & interconnect	Copper direct-attach cables	$1.05B	0.35%	+32.0%	Medium	Copper remains useful at short reaches, though optics and active cables take more of the premium mix.	S3 S5 S13
82	Accelerators & compute	Accelerator respins and yield buffer	$1.04B	0.35%	+25.0%	Low	Respins and yield buffers rise when ramps are messy, but they are not the main event once volume stabilizes.	S1 S2 S3 S4 S12
83	HBM	Long-tail custom HBM configurations	$0.99B	0.33%	+110.0%	Low	Custom or unusual stack configurations matter at the margin, not at center scale.	S1 S2 S6 S7 S12
84	Advanced packaging, substrates & test	Chip-on-wafer bonding materials	$0.98B	0.33%	+30.0%	Medium	Bonding materials scale with advanced-packaging throughput.	S2 S7 S12
85	Networking & interconnect	Active electrical cables	$0.97B	0.32%	+42.0%	Medium	AECs grow with rack density but remain below transceiver-class dollar intensity.	S3 S5 S13
86	Optics & photonics	Silicon photonics engines	$0.94B	0.32%	+70.0%	Medium	Integrated photonics gains value as bandwidth density and power efficiency matter more.	S5 S13
87	Standard memory	Legacy DDR4 support inventory	$0.89B	0.30%	-10.0%	Low	Legacy memory stays in service where platforms lag, but strategic spend is moving to newer standards.	S2 S6 S12
88	Networking & interconnect	Backplane and midplane interconnects	$0.83B	0.28%	+36.0%	Low	Backplane complexity increases, but this spend grows slower than active networking silicon.	S3 S5 S13
89	Advanced packaging, substrates & test	Advanced package assembly	$0.82B	0.28%	+25.0%	Low	Assembly volume grows, but pricing power is lower than bottleneck services.	S2 S7 S12
90	Storage	Spare and replacement drives	$0.72B	0.24%	+8.0%	Low	Spare pools are operationally important but not a strategic growth bucket.	S8 S9 S10 S11
91	Accelerators & compute	Long-tail specialty accelerators	$0.69B	0.23%	+18.0%	Low	Specialized or niche AI accelerators exist, but broad hyperscaler demand concentrates value elsewhere.	S1 S2 S3 S4 S12
92	Optics & photonics	Optical couplers / waveguides	$0.67B	0.23%	+40.0%	Low	Passive optical components expand with system complexity but stay a secondary spend line.	S5 S13
93	Advanced packaging, substrates & test	Thermal lids and integrated heat spreaders	$0.66B	0.22%	+20.0%	Low	Thermal hardware scales with package density but is not the core value driver.	S2 S7 S12
94	Standard memory	Long-tail specialty memory modules	$0.63B	0.21%	+17.0%	Low	Niche modules exist, though most value is concentrated in mainstream server memory formats.	S2 S6 S12
95	Networking & interconnect	Timing / clock ICs for AI fabrics	$0.55B	0.19%	+20.0%	Low	Timing components are critical but small in value relative to headline fabric silicon.	S3 S5 S13
96	Storage	Archive / object-storage media	$0.54B	0.18%	+12.0%	Low	Archival layers remain necessary, though faster storage tiers capture more current spend.	S8 S9 S10 S11
97	Advanced packaging, substrates & test	Yield-learning / rework capacity	$0.41B	0.14%	+10.0%	Low	Rework and learning capacity matter most during ramps, then become a smaller share.	S2 S7 S12
98	Optics & photonics	Fiber connectivity and patching	$0.40B	0.14%	+32.0%	Low	Physical fiber plant is necessary but captures less value than active optical components.	S5 S13
99	Optics & photonics	Next-gen 3.2T optical readiness inventory	$0.27B	0.09%	+100.0%	Low	Readiness spending is still small, but the next speed transition is already pulling forward work.	S5 S13
100	Networking & interconnect	Long-tail network spares and buffers	$0.19B	0.06%	+10.0%	Low	Long-tail support items matter operationally, not strategically.	S3 S5 S13

Direct comparison: carbon wants shelter, silicon wants throughput

This is the section that translates the spreadsheet work back into the tweet's intuition.

	Human / carbon-based basket	Silicon / AI-cluster basket
Accounting basis	Official U.S. 2025 personal consumption expenditures, from BEA/FRED. H3	Modeled 2025 AI-server-related procurement basket anchored to TrendForce, cross-checked with IDC server and networking data. S1 S2 S3 S5
What dominates	Shelter, health care, food, finance, and broad services. H3	Accelerators, HBM, networking/interconnect, host CPUs, storage, and advanced packaging. S1 S2 S3 S7
Buyer pattern	Diffuse: nearly every household, every week, across millions of independent decisions.	Concentrated: hyperscalers, cloud platforms, sovereign projects, and a narrow enterprise buyer set.
Growth regime	Mostly mid-single-digit YoY at the top level, with occasional commodity-price swings. H3	High-double-digit to triple-digit in key bottlenecks such as accelerators, networking, and HBM. S2 S5
Main constraint	Income, demographics, rent, healthcare intensity, and service labor availability.	Advanced packaging, HBM supply, leading-edge foundry capacity, power density, and interconnect bandwidth. S2 S7 S13
Interpretation	Human consumption optimizes for survival, comfort, mobility, and social life.	Silicon consumption optimizes for tokens, training throughput, memory bandwidth, and latency.

Human basket in one line

Human spending is diffuse, repetitive, and necessity-loaded. The core functions are: shelter, health, food, mobility, finance, and social participation. H3

Silicon basket in one line

Silicon spending is concentrated, capex-led, and bottleneck-loaded. The core functions are: compute, memory bandwidth, network bandwidth, packaging yield, storage latency, and power density. S1 S2 S3 S5 S7

Best way to read the analogy

The right interpretation is not “AI literally consumes like humans.” It is “capital markets are increasingly prioritizing the hardware stack that keeps model systems alive, fed, and scaling.” S1 S2 S3

Stock ideas by basket

These are exposure proxies chosen to map onto the basket logic, not perfect one-to-one statistical matches.

Human / carbon-based basket — 2 U.S. + 2 non-U.S.

Walmart

NYSE: WMT

US-based

Scale retail, grocery, consumables, pharmacy

FY2026 revenue: $713B

A broad proxy for recurring household spend, especially groceries, everyday goods, and pharmacy-led traffic.

Risk watch: Low-margin retail; mix shifts and consumer trade-down can change profit capture even if basket demand stays firm.

Eli Lilly

NYSE: LLY

US-based

Pharmaceuticals

FY2025 revenue: $65.2B

Direct proxy for one of the biggest current human-goods lines: pharmaceutical and medical-product consumption.

Risk watch: Concentration in a few blockbuster franchises and policy risk around drug pricing.

Nestlé

SIX: NESN

Non-US-based

Packaged food, beverages, nutrition, pet care

2025 sales: CHF 89.5B

A diversified way to express everyday food-and-beverage demand, with additional exposure to nutrition and pet categories.

Risk watch: FX, coffee/cocoa costs, and slower discretionary food categories.

Novartis

SIX: NOVN / NYSE: NVS

Non-US-based

Innovative pharmaceuticals

FY2025 net sales: $54.5B

Another direct proxy for the large and growing pharmaceutical component of the human consumption basket.

Risk watch: Patent cliffs, pipeline execution, and reimbursement pressure.

Silicon / AI-cluster basket — 2 U.S. + 2 non-U.S.

NVIDIA

NASDAQ: NVDA

US-based

Accelerators, AI systems, networking

FY2026 revenue: $215.9B; Data Center: $193.7B

The clearest listed proxy for the single largest silicon-basket bucket: high-end accelerators and AI-system demand.

Risk watch: Customer concentration, export controls, and the pace of custom-ASIC substitution.

Broadcom

NASDAQ: AVGO

US-based

Custom AI accelerators, AI networking, connectivity

Q1 FY2026 AI revenue: $8.4B (+106% YoY)

Strong exposure to two fast-growing silicon needs at once: custom accelerators and AI networking/interconnect.

Risk watch: AI program concentration among a small number of hyperscale buyers.

C6 C10

TSMC

TWSE: 2330 / NYSE ADR: TSM

Non-US-based

Leading-edge foundry

2025 revenue: NT$3,809.05B; advanced nodes 77% of Q4 wafer revenue

A core manufacturing bottleneck for the compute-heavy side of the silicon basket, especially leading-edge AI chips.

Risk watch: Geopolitics, capex intensity, and cyclicality outside AI.

C7 C8

SK hynix

KRX: 000660

Non-US-based

HBM and AI memory

FY2025 revenue: 97.1467T won

Direct exposure to the fastest-growing memory layer in the silicon basket: HBM attached to AI accelerators.

Risk watch: Memory-price cyclicality and execution on next-generation HBM qualification.

Open the stock recommendation table

Stock ideas by consumption type

Informational only; these are exposure proxies, not personalized investment advice.

Theme ↕	Region ↕	Company ↕	Ticker ↕	Exposure ↕	Headline fact ↕	Why it fits ↕	Risk watch ↕	Sources ↕
Human / carbon-based consumption	US-based	Walmart	NYSE: WMT	Scale retail, grocery, consumables, pharmacy	FY2026 revenue: $713B	A broad proxy for recurring household spend, especially groceries, everyday goods, and pharmacy-led traffic.	Low-margin retail; mix shifts and consumer trade-down can change profit capture even if basket demand stays firm.	C1
Human / carbon-based consumption	US-based	Eli Lilly	NYSE: LLY	Pharmaceuticals	FY2025 revenue: $65.2B	Direct proxy for one of the biggest current human-goods lines: pharmaceutical and medical-product consumption.	Concentration in a few blockbuster franchises and policy risk around drug pricing.	C2
Human / carbon-based consumption	Non-US-based	Nestlé	SIX: NESN	Packaged food, beverages, nutrition, pet care	2025 sales: CHF 89.5B	A diversified way to express everyday food-and-beverage demand, with additional exposure to nutrition and pet categories.	FX, coffee/cocoa costs, and slower discretionary food categories.	C3
Human / carbon-based consumption	Non-US-based	Novartis	SIX: NOVN / NYSE: NVS	Innovative pharmaceuticals	FY2025 net sales: $54.5B	Another direct proxy for the large and growing pharmaceutical component of the human consumption basket.	Patent cliffs, pipeline execution, and reimbursement pressure.	C4
Silicon / AI-cluster consumption	US-based	NVIDIA	NASDAQ: NVDA	Accelerators, AI systems, networking	FY2026 revenue: $215.9B; Data Center: $193.7B	The clearest listed proxy for the single largest silicon-basket bucket: high-end accelerators and AI-system demand.	Customer concentration, export controls, and the pace of custom-ASIC substitution.	C5
Silicon / AI-cluster consumption	US-based	Broadcom	NASDAQ: AVGO	Custom AI accelerators, AI networking, connectivity	Q1 FY2026 AI revenue: $8.4B (+106% YoY)	Strong exposure to two fast-growing silicon needs at once: custom accelerators and AI networking/interconnect.	AI program concentration among a small number of hyperscale buyers.	C6 C10
Silicon / AI-cluster consumption	Non-US-based	TSMC	TWSE: 2330 / NYSE ADR: TSM	Leading-edge foundry	2025 revenue: NT$3,809.05B; advanced nodes 77% of Q4 wafer revenue	A core manufacturing bottleneck for the compute-heavy side of the silicon basket, especially leading-edge AI chips.	Geopolitics, capex intensity, and cyclicality outside AI.	C7 C8
Silicon / AI-cluster consumption	Non-US-based	SK hynix	KRX: 000660	HBM and AI memory	FY2025 revenue: 97.1467T won	Direct exposure to the fastest-growing memory layer in the silicon basket: HBM attached to AI accelerators.	Memory-price cyclicality and execution on next-generation HBM qualification.	C9

Downloads

Flat files used to build the page. If you want to re-rank, filter, or chart the data yourself, start here.

human_published_100.csv human_clean_partition_16.csv human_top_goods_leaf_20.csv silicon_modeled_100.csv silicon_group_summary.csv stock_recommendations.csv sources.csv

Complete source appendix

Every source used in the page, including the uploaded tweet screenshot and all web references.

Source cards

H0 Tweet screenshot (uploaded by user)

local-upload:/mnt/data/Screenshot 2026-04-20 at 11.36.20 AM.png

Project inspiration and the framing quote.

H1 BEA Consumer Spending page

https://www.bea.gov/index.php/data/consumer-spending/main

Methodology note that BEA underlying detail tables exist and that detailed estimates are less reliable than higher aggregates.

H2 BEA FAQ 84 on detailed PCE estimates and the PCE Bridge

https://www.bea.gov/help/faq/84

Methodological limit: Table 2.4.5U is the most detailed current time-series PCE table; benchmark-year bridge provides more commodity detail.

H3 FRED / BEA Table 2.4.5 Personal Consumption Expenditures by Type of Product (2025 annual)

https://fred.stlouisfed.org/release/tables?eid=44183&rid=53

All 2025 and 2024 human consumption amounts in the page, including the top-100 current published series.

S1 TrendForce press-center listing for Jan. 6, 2025 AI server outlook

https://www.trendforce.com/presscenter/news/Semiconductors?page=21

Anchor for the modeled silicon basket: AI-server-related industry value of US$298B in 2025 and AI servers above 70% of server-industry value.

S2 TrendForce: Strong Demand from CSPs and Sovereign Cloud to Drive Over 20% Growth in AI Server Shipments by 2026

https://www.trendforce.com/presscenter/news/20251030-12762.html

2025 AI server revenue almost +48%, 2026 AI servers 74% of server-market value, NVIDIA ~70% AI-chip share in 2025, and HBM demand +130% from 2025 AI-chip shipments.

S3 IDC 4Q25 Worldwide Server Tracker press release

https://www.idc.com/resource-center/press-releases/4q25servertracker/

Official 2025 global server market result ($444.1B, +80.4%) plus IDC server-system taxonomy.

S4 IDC Servers Market Insights

https://www.idc.com/promo/servers/

Context that AI infrastructure buildout remained the dominant force in late 2025.

S5 IDC Ethernet switch market size and growth blog

https://www.idc.com/resource-center/blog/ethernet-switch-market-size-and-growth-datacenter-segment-surges-60-in-q4-as-ai-workloads-expand/

Network-growth signal: datacenter Ethernet switch market growth and 800G mix.

S6 TrendForce memory revenue outlook (HBM / DRAM / NAND)

https://www.trendforce.com/presscenter/news/20240722-12228.html

Memory-market context, including DRAM 2025 revenue outlook and the role of HBM.

S7 SEMI 2025 semiconductor equipment billings

https://www.semi.org/en/SEMI-Reports-Global-Semiconductor-Equipment-Billings-Reached-135-Billion-in-2025

Packaging/test capacity signal: assembly & packaging +21%, test +55% in 2025.

S8 TrendForce 2Q25 enterprise SSD revenue update

https://www.trendforce.com/presscenter/news/20250908-12703.html

Storage demand signal: top-five enterprise SSD revenue +12.7% QoQ in 2Q25.

S9 TrendForce 3Q25 enterprise SSD revenue update

https://www.trendforce.com/presscenter/news/20251205-12819.html

Storage demand signal: top-five enterprise SSD revenue +28% in 3Q25.

S10 TrendForce 3Q25 NAND-flash demand update

https://www.trendforce.com/presscenter/news/20251203-12813.html

NAND demand context from continuing AI-infrastructure buildout.

S11 TrendForce nearline HDD shortages / QLC SSD breakout note

https://www.trendforce.com/presscenter/news/20250915-12714.html

HDD supply stress and QLC SSD adoption signal.

S12 SIA global semiconductor sales 2025

https://www.semiconductors.org/global-annual-semiconductor-sales-increase-25-6-to-791-7-billion-in-2025/

Semiconductor macro context: 2025 sales total and logic/memory leadership.

S13 Broadcom OFC 2026 AI infrastructure portfolio

https://investors.broadcom.com/news-releases/news-release-details/broadcom-showcases-industry-leading-solutions-scaling-ai

Examples of live AI-infrastructure goods: 102.4T switch, 800G NIC, optics, PCIe, co-packaged optics.

C1 Walmart FY2026 revenue / company overview note

https://corporate.walmart.com/news/2026/02/19/walmart-raises-annual-dividend-to-0-99-per-share-marking-53rd-consecutive-year-of-dividend-increases

Walmart revenue scale and company profile.

C2 Eli Lilly fourth-quarter and full-year 2025 results

https://investor.lilly.com/news-releases/news-release-details/lilly-reports-fourth-quarter-2025-financial-results-and-provides

Lilly revenue scale and pharma exposure.

C3 Nestlé full-year 2025 results

https://www.nestle.com/media/pressreleases/allpressreleases/full-year-results-2025

Nestlé revenue scale and category exposure.

C4 Novartis annual results 2025

https://www.novartis.com/Investors/financial-data/Annual-results

Novartis revenue scale and pharma exposure.

C5 NVIDIA fiscal 2026 results

https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-fourth-quarter-and-fiscal-2026

NVIDIA revenue and data-center exposure.

C6 Broadcom Q1 FY2026 results

https://investors.broadcom.com/node/63976/pdf

Broadcom AI revenue and custom-accelerator / networking exposure.

C7 TSMC December 2025 revenue report

https://pr.tsmc.com/english/news/3278

TSMC 2025 revenue scale.

C8 TSMC Q4 2025 results

https://pr.tsmc.com/english/news/3281

TSMC advanced-node mix (77% of Q4 wafer revenue).

C9 SK hynix FY2025 results

https://news.skhynix.com/sk-hynix-announces-fy25-financial-results/

SK hynix revenue scale and AI-memory/HBM exposure.

C10 Broadcom OFC 2026 AI infrastructure portfolio

https://investors.broadcom.com/news-releases/news-release-details/broadcom-showcases-industry-leading-solutions-scaling-ai

Broadcom product-level AI-infrastructure examples (switches, NICs, optics, PCIe).

Open the same source list as a searchable table

Complete source appendix

Every source used in the page is listed here. H0 is the uploaded tweet screenshot; everything else is a web source.

Tag ↕	Source ↕	URL / location ↕	Used for ↕
H0	Tweet screenshot (uploaded by user)	`local-upload:/mnt/data/Screenshot 2026-04-20 at 11.36.20 AM.png`	Project inspiration and the framing quote.
H1	BEA Consumer Spending page	`https://www.bea.gov/index.php/data/consumer-spending/main`	Methodology note that BEA underlying detail tables exist and that detailed estimates are less reliable than higher aggregates.
H2	BEA FAQ 84 on detailed PCE estimates and the PCE Bridge	`https://www.bea.gov/help/faq/84`	Methodological limit: Table 2.4.5U is the most detailed current time-series PCE table; benchmark-year bridge provides more commodity detail.
H3	FRED / BEA Table 2.4.5 Personal Consumption Expenditures by Type of Product (2025 annual)	`https://fred.stlouisfed.org/release/tables?eid=44183&rid=53`	All 2025 and 2024 human consumption amounts in the page, including the top-100 current published series.
S1	TrendForce press-center listing for Jan. 6, 2025 AI server outlook	`https://www.trendforce.com/presscenter/news/Semiconductors?page=21`	Anchor for the modeled silicon basket: AI-server-related industry value of US$298B in 2025 and AI servers above 70% of server-industry value.
S2	TrendForce: Strong Demand from CSPs and Sovereign Cloud to Drive Over 20% Growth in AI Server Shipments by 2026	`https://www.trendforce.com/presscenter/news/20251030-12762.html`	2025 AI server revenue almost +48%, 2026 AI servers 74% of server-market value, NVIDIA ~70% AI-chip share in 2025, and HBM demand +130% from 2025 AI-chip shipments.
S3	IDC 4Q25 Worldwide Server Tracker press release	`https://www.idc.com/resource-center/press-releases/4q25servertracker/`	Official 2025 global server market result ($444.1B, +80.4%) plus IDC server-system taxonomy.
S4	IDC Servers Market Insights	`https://www.idc.com/promo/servers/`	Context that AI infrastructure buildout remained the dominant force in late 2025.
S5	IDC Ethernet switch market size and growth blog	`https://www.idc.com/resource-center/blog/ethernet-switch-market-size-and-growth-datacenter-segment-surges-60-in-q4-as-ai-workloads-expand/`	Network-growth signal: datacenter Ethernet switch market growth and 800G mix.
S6	TrendForce memory revenue outlook (HBM / DRAM / NAND)	`https://www.trendforce.com/presscenter/news/20240722-12228.html`	Memory-market context, including DRAM 2025 revenue outlook and the role of HBM.
S7	SEMI 2025 semiconductor equipment billings	`https://www.semi.org/en/SEMI-Reports-Global-Semiconductor-Equipment-Billings-Reached-135-Billion-in-2025`	Packaging/test capacity signal: assembly & packaging +21%, test +55% in 2025.
S8	TrendForce 2Q25 enterprise SSD revenue update	`https://www.trendforce.com/presscenter/news/20250908-12703.html`	Storage demand signal: top-five enterprise SSD revenue +12.7% QoQ in 2Q25.
S9	TrendForce 3Q25 enterprise SSD revenue update	`https://www.trendforce.com/presscenter/news/20251205-12819.html`	Storage demand signal: top-five enterprise SSD revenue +28% in 3Q25.
S10	TrendForce 3Q25 NAND-flash demand update	`https://www.trendforce.com/presscenter/news/20251203-12813.html`	NAND demand context from continuing AI-infrastructure buildout.
S11	TrendForce nearline HDD shortages / QLC SSD breakout note	`https://www.trendforce.com/presscenter/news/20250915-12714.html`	HDD supply stress and QLC SSD adoption signal.
S12	SIA global semiconductor sales 2025	`https://www.semiconductors.org/global-annual-semiconductor-sales-increase-25-6-to-791-7-billion-in-2025/`	Semiconductor macro context: 2025 sales total and logic/memory leadership.
S13	Broadcom OFC 2026 AI infrastructure portfolio	`https://investors.broadcom.com/news-releases/news-release-details/broadcom-showcases-industry-leading-solutions-scaling-ai`	Examples of live AI-infrastructure goods: 102.4T switch, 800G NIC, optics, PCIe, co-packaged optics.
C1	Walmart FY2026 revenue / company overview note	`https://corporate.walmart.com/news/2026/02/19/walmart-raises-annual-dividend-to-0-99-per-share-marking-53rd-consecutive-year-of-dividend-increases`	Walmart revenue scale and company profile.
C2	Eli Lilly fourth-quarter and full-year 2025 results	`https://investor.lilly.com/news-releases/news-release-details/lilly-reports-fourth-quarter-2025-financial-results-and-provides`	Lilly revenue scale and pharma exposure.
C3	Nestlé full-year 2025 results	`https://www.nestle.com/media/pressreleases/allpressreleases/full-year-results-2025`	Nestlé revenue scale and category exposure.
C4	Novartis annual results 2025	`https://www.novartis.com/Investors/financial-data/Annual-results`	Novartis revenue scale and pharma exposure.
C5	NVIDIA fiscal 2026 results	`https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-fourth-quarter-and-fiscal-2026`	NVIDIA revenue and data-center exposure.
C6	Broadcom Q1 FY2026 results	`https://investors.broadcom.com/node/63976/pdf`	Broadcom AI revenue and custom-accelerator / networking exposure.
C7	TSMC December 2025 revenue report	`https://pr.tsmc.com/english/news/3278`	TSMC 2025 revenue scale.
C8	TSMC Q4 2025 results	`https://pr.tsmc.com/english/news/3281`	TSMC advanced-node mix (77% of Q4 wafer revenue).
C9	SK hynix FY2025 results	`https://news.skhynix.com/sk-hynix-announces-fy25-financial-results/`	SK hynix revenue scale and AI-memory/HBM exposure.
C10	Broadcom OFC 2026 AI infrastructure portfolio	`https://investors.broadcom.com/news-releases/news-release-details/broadcom-showcases-industry-leading-solutions-scaling-ai`	Broadcom product-level AI-infrastructure examples (switches, NICs, optics, PCIe).