No one agrees how much AI spending reaches chips: about 25% in one Goldman-based analysis, 50% in BNP Paribas, about 70% in PwC. They measure different things, so every chip share outside company statements is an estimate.
Back to AI Council · Markets: Markets today, 2026-11-07: fear index (VIX) −9.0%, chip stocks (SOX) +4.2%, Nasdaq-100 +1.4% All Markets ratings →
Poly Rating · from AI Council · Markets
Who has the most AI chips? Top 10 AI spenders of 2026
Amazon has the biggest 2026 budget at about $220 billion, but of the big four only Alphabet and Microsoft say how much goes to hardware: about 65% and about three-quarters of their latest quarter.
- No. 1: Amazon (AWS)
- 10 places
01Final list
The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.
Order: from most to least AI agreement.
A bright badge means this AI put the entry on its list. A dim badge means it did not.
- 1
Amazon (AWS)
Biggest budget at about $220B, stated in its filing, but no published chip split, so it sits behind two companies that show theirs.
Included it: Claude, Meta, Grok, Gemini, Nova, Kimi, DeepSeek, Qwen. Did not include: MiMo, Z.ai.In their lists: Claude: 1 · Meta: 2 · Grok: 1 · Gemini: 4 · Nova: 2 · Kimi: 2 · DeepSeek: 2 · Qwen: 5
What each AI noted 8
- Claudeplace 1
The largest 2026 budget, about $220B, raised from $200B in July. Management says 2027 will be 'meaningfully more'.
Source ↗ - Metaplace 2
Largest headline budget at $220B, with a $578B AWS backlog that makes the chip bill look like a down payment.
Source ↗ - Grokplace 1
~$220B cash CapEx guide, Q3 alone $77B, still the largest absolute pipeline into GPUs and Trainium
Source ↗ - Geminiplace 4
Boasts the world's largest $210B total capex envelope, though heavy civil real estate and logistics keep chip conversion closer to 51%.
- Novaplace 2
$220B capex where two-thirds ($145B) targets AI compute – logistics drag limits pure chip focus
Source ↗ - Kimiplace 2
Reiterated ~$220B for 2026 with 2027 'meaningfully more' — and AWS growing 39%, its fastest in five years, means the chips are spoken for.
Source ↗ - DeepSeekplace 2
The $220 billion budget grew by $20 billion in July for one reason — memory prices — the cleanest proof on this list that the chip line, not the building line, is what moves.
Source ↗ - Qwenplace 5
Raised its budget by $20B mostly to absorb soaring memory prices for custom Trainium and Nvidia racks.
Source ↗
8 AIs put it here - 2
Microsoft
Highest disclosed chip share, about 3/4 of its latest quarter and 2/3 the quarter before, but its budget was trimmed to $170B after outside capacity arrived late.
Included it: Claude, Meta, Grok, Gemini, Nova, Kimi, DeepSeek, Qwen. Did not include: MiMo, Z.ai.In their lists: Claude: 2 · Meta: 3 · Grok: 3 · Gemini: 1 · Nova: 3 · Kimi: 1 · DeepSeek: 3 · Qwen: 7
What each AI noted 8
- Claudeplace 2
It cut 2026 capex to $170B from $175B and still rose after earnings. Its CFO says spending accelerates in H1 2027 because components cost more.
Source ↗ - Metaplace 3
The only big spender the market rewarded after earnings: Azure up 48% while capex came in below its own $50B-plus plan.
Source ↗ - Grokplace 3
~ $170B calendar guide, ~75% short-lived assets (CPUs/GPUs) — highest disclosed chip intensity
Source ↗ - Geminiplace 1
Funneling an industry-leading $115 billion straight into silicon across Nvidia Blackwell clusters and custom Maia accelerators.
- Novaplace 3
$170B capex with 66% ($112B) on GPUs/CPUs – Azure’s infrastructure play
Source ↗ - Kimiplace 1
Guides ~$255-260B for FY2027 and says two-thirds of every capex dollar already goes to 'short-lived' assets — mostly CPUs and GPUs.
Source ↗ - DeepSeekplace 3
Wall Street wanted $154.6 billion; Microsoft guided $190 billion, and $25 billion of the difference is memory and component costs — a capex beat made of chips.
Source ↗ - Qwenplace 7
Cut its forecast to $170B as supply chain delays force it to extend the useful life of older servers.
Source ↗
8 AIs put it here - 3
Alphabet (Google)
Largest disclosed hardware line: about 65% of Q3 infrastructure spending went to servers and networking, on a $195-205B budget. Its stock still fell about 12%.
Included it: Claude, Meta, Grok, Gemini, Nova, Kimi, DeepSeek, Qwen. Did not include: MiMo, Z.ai.In their lists: Claude: 3 · Meta: 4 · Grok: 2 · Gemini: 3 · Nova: 1 · Kimi: 3 · DeepSeek: 1 · Qwen: 6
What each AI noted 8
- Claudeplace 3
It held a $195-205B range and the stock fell about 10-12%, the sharpest punishment of any spender this season.
Source ↗ - Metaplace 4
$195B-$205B guided, but 65% went to servers in Q3 and TPUs now ship in 78% of its AI servers.
Source ↗ - Grokplace 2
$195-205B range held, 65% of technical infrastructure already servers and networking
Source ↗ - Geminiplace 3
Splits its $110B hardware outlay evenly between commercial Nvidia GPUs and custom in-house TPU v5p/v6e silicon co-designed with Broadcom.
- Novaplace 1
$185B total capex with $130B directly tied to AI chips – the clearest chip focus among giants
- Kimiplace 3
Doubled its capex plan in 18 months to $195-205B — and even selling its own TPUs to cloud customers now counts as chip demand.
Source ↗ - DeepSeekplace 1
The $205 billion budget is the largest single chip-directed pool on this list: about 60% goes to servers, roughly $123 billion, before a single dollar of buildings or power.
Source ↗ - Qwenplace 6
Guiding $205B but taking a 10% stock hit as investors fear the cash is leaking into power plants instead of TPUs.
Source ↗
8 AIs put it here - 4
Meta Platforms
$135-145B with no published split; a multi-year Nvidia deal and a $60B AMD deal point to chips. The stock fell about 7% the day after it raised its floor.
Included it: Claude, Meta, Grok, Gemini, Nova, Kimi, DeepSeek, Qwen. Did not include: MiMo, Z.ai.In their lists: Claude: 4 · Meta: 1 · Grok: 4 · Gemini: 2 · Nova: 4 · Kimi: 4 · DeepSeek: 6 · Qwen: 4
What each AI noted 8
- Claudeplace 4
It raised only the floor of its range ($130B to $135B) and still lost 3% on the first reaction, then rose 3% a week later.
Source ↗ - Metaplace 1
The only spender with signed chip deals bigger than most rivals' whole budgets: about $50B Nvidia plus $60B AMD.
Source ↗ - Grokplace 4
$135-145B range, floor raised, heavy Llama/superintelligence server demand
Source ↗ - Geminiplace 2
Directs 68% of its budget into pure compute silicon, avoiding third-party commercial cloud overhead to train open-source Llama models.
- Novaplace 4
$145B capex where 44% ($64B) funds custom AI silicon – partnership-driven acceleration
Source ↗ - Kimiplace 4
Raised its capex floor four quarters in a row to $135-145B and is still 'supply constrained on compute' — so it rents even more chips from neoclouds on top.
Source ↗ - DeepSeekplace 6
Meta's $145 billion budget is the one where chip suppliers are literally buying the equity — Nvidia, AMD and Broadcom offered up to $5 billion — because the building and power bill is so large.
Source ↗ - Qwenplace 4
Back on offense with a $145B ceiling that prioritizes training Llama over traditional ad infrastructure.
Source ↗
8 AIs put it here - 5
Oracle
Fiscal 2027 budget is $90-95B gross but at most $70B net cash, with customers funding the rest. One source puts over 60% on GPU clusters (estimate).
Included it: Claude, Meta, Grok, Gemini, Nova, Kimi, DeepSeek, Qwen. Did not include: MiMo, Z.ai.In their lists: Claude: 5 · Meta: 6 · Grok: 5 · Gemini: 5 · Nova: 5 · Kimi: 5 · DeepSeek: 4 · Qwen: 8
What each AI noted 8
- Claudeplace 5
Fiscal 2027 gross capex of $90-95B, but its CFO says net cash capex is 'not more than $70B'. Remaining performance obligations are $664B.
- Metaplace 6
Gross $90B-$95B sounds huge, but net cash is capped at $70B: $20B-$25B is customer prepayments, not Oracle cash.
- Grokplace 5
$90-95B fiscal guidance, pure cloud/AI capacity build with large GPU deployments
- Geminiplace 5
Transforms 65% of its surging $55B budget directly into high-density GPU bare-metal clusters dedicated to leading frontier model builders.
- Novaplace 5
$58.8B capex with 40% ($23.5B) on cloud AI infrastructure – database-to-accelerator pivot
Source ↗ - Kimiplace 5
Guides $90-95B for FY2027 — but $75B of its AI contracts are customer-prepaid or customer-supplied GPUs, so the chips arrive even when Oracle's cash doesn't pay for them.
Source ↗ - DeepSeekplace 4
A $95 billion reported capex budget against roughly $67.7 billion of analyst expectations, with more than 60% aimed at Nvidia GPU clusters — and the stock paid for it with a double-digit drop.
Source ↗ - Qwenplace 8
Surging 162% to build cloud regions, but weighed down by $130B in debt to finance the concrete.
8 AIs put it here - 6
CoreWeave
$35-39B, nearly all GPUs (estimate, no published split), against $12.4-13.2B of revenue: the purest chip buyer here, on a far smaller budget.
Included it: Claude, Meta, Grok, Z.ai, Kimi, DeepSeek, Qwen. Did not include: Gemini, Nova, MiMo.In their lists: Claude: 6 · Meta: 5 · Grok: 6 · Z.ai: 7 · Kimi: 7 · DeepSeek: 5 · Qwen: 1
What each AI noted 7
- Claudeplace 6
Its $35-39B budget is small, but it is nearly all GPUs and the buildings are leased.
Source ↗ - Metaplace 5
The purest GPU landlord: $35B-$39B of capex against only $12.4B-$13.2B of revenue guidance.
Source ↗ - Grokplace 6
$31-39B pure-play GPU cloud spend — almost all dollars flow straight to chips
Source ↗ - Z.aiplace 7
Capex of $35-39 billion against guided revenue of $12.4-13.2 billion - the purest GPU-per-dollar ratio in the group, with a ~$104 billion backlog behind it.
Source ↗ - Kimiplace 7
The purest chip buyer on the list: ~$37B of capex that is almost entirely GPUs, with a $104B backlog pre-paying for them.
Source ↗ - DeepSeekplace 5
It guides $35–$39 billion of capex on $12.4–$13.2 billion of revenue — a capex-to-revenue ratio above 3x, the highest purity of chip spending on this list.
Source ↗ - Qwenplace 1
A pure GPU proxy with a $104B backlog that converts debt almost entirely into Nvidia orders.
Source ↗
7 AIs put it here - 7
Alibaba
June-quarter spending hit RMB67.7B, up 75%, with half its RMB380B plan spent. CLSA's full-year RMB220B (about $33B) is an estimate, and its chip share is unpublished.
Included it: Claude, Meta, Gemini, Nova, Z.ai, Kimi, DeepSeek. Did not include: Grok, MiMo, Qwen.In their lists: Claude: 7 · Meta: 8 · Gemini: 7 · Nova: 6 · Z.ai: 8 · Kimi: 6 · DeepSeek: 8
What each AI noted 7
- Claudeplace 7
The biggest Chinese spender: RMB190B of a RMB380B three-year plan is spent, and June-quarter capex rose 75%.
Source ↗ - Metaplace 8
The smallest of the China giants on paper, but one June quarter alone was RMB 67.7B: 53.7% of its whole prior fiscal year.
Source ↗ - Geminiplace 7
Ramps cloud capital spending by over 30% to feed proprietary Qwen models, deploying both international silicon and domestic Chinese accelerators.
- Novaplace 6
$126.9B capex but opaque chip split – domestic China dynamics limit transparency
- Z.aiplace 8
Half of its RMB 380 billion three-year plan is already spent; June-quarter capex rose 75% while net income fell 75% - and the stock is down 25% year to date anyway.
Source ↗ - Kimiplace 6
Halfway through its RMB 380B ($56B) three-year AI plan after 17 months — and already says it will exceed it, while profit fell 99.7% to pay for it.
Source ↗ - DeepSeekplace 8
CLSA sees RMB220 billion, about $33 billion, of 2026 capex — China's largest listed AI capex line that visibly reaches chipmakers, even as the ADR stays under pressure.
Source ↗
7 AIs put it here - 8
Tesla
Cut to about $20B from more than $25B, and factories, robots and robotaxis share the budget, so chips are reported as a minority, with no published split.
Included it: Meta, Grok, Gemini, Kimi, DeepSeek, Qwen. Did not include: Claude, Nova, MiMo, Z.ai.In their lists: Meta: 10 · Grok: 7 · Gemini: 6 · Kimi: 9 · DeepSeek: 7 · Qwen: 9
What each AI noted 6
- Metaplace 10
Cut its guide instead of raising it: $20B, down from $25B, while Q3 capex still doubled to $4.7B.
Source ↗ - Grokplace 7
$20B guide, Dojo and AI inference silicon plus data-center build
Source ↗ - Geminiplace 6
Dedicates $12.4B straight to Nvidia hardware and custom AI chips for Cortex training superclusters and autonomous driving fleets.
- Kimiplace 9
The only top-10 name that CUT — $20B, down from 'more than $25B' — and much of it buys factories and robotaxis, not chips.
Source ↗ - DeepSeekplace 7
A $25 billion capex year with close to $20 billion tagged for AI — but the stock fell 14.5%, the sharpest rejection of any budget on this list.
Source ↗ - Qwenplace 9
Slashing its budget to $20B to focus on Dojo and factories rather than standard cloud AI chips.
Source ↗
6 AIs put it here - 9
Tencent
No full-year guide. Q2 hit RMB52.8B, up 176%, and 2026 estimates run from RMB181B to RMB250B; most AI investment from a $5B bond is reported as buying chips.
Included it: Claude, Meta, Gemini, Nova, Z.ai, DeepSeek. Did not include: Grok, MiMo, Kimi, Qwen.In their lists: Claude: 10 · Meta: 7 · Gemini: 8 · Nova: 8 · Z.ai: 9 · DeepSeek: 10
What each AI noted 6
- Claudeplace 10
China's cloud trio adds roughly $35B of spending beyond Alibaba, but no company-level guidance surfaced.
- Metaplace 7
Q2 alone was RMB 52.8B, up 176% year on year: at that pace Tencent is a $29B-plus annual spender without ever giving a formal full-year guide.
Source ↗ - Geminiplace 8
Tripled quarterly infrastructure run-rates to secure high-bandwidth memory and training accelerators for its multimodal Hunyuan model.
- Novaplace 8
HK-listed $54.6B capex with growing AI cloud share – China’s silent infrastructure builder
Source ↗ - Z.aiplace 9
Operating capex jumped 190% in a single quarter to RMB 51.8 billion - and with US chip curbs in force, a growing slice flows to domestic silicon instead of Nvidia.
- DeepSeekplace 10
Tencent's AI capex more than doubled to RMB96.3 billion in H1 2026, and it is borrowing $5 billion mainly to buy chips — yet the stock hit a one-year low.
Source ↗
6 AIs put it here - 10
Nebius
$20-25B guide, raised from $16-20B, against only $3-3.4B of revenue. Mostly GPUs (estimate), though data centres sit in the same spending line.
Included it: Claude, Meta, Grok, Z.ai, Kimi. Did not include: Gemini, Nova, MiMo, DeepSeek, Qwen.In their lists: Claude: 9 · Meta: 9 · Grok: 8 · Z.ai: 10 · Kimi: 8
What each AI noted 5
- Claudeplace 9
Combined with CoreWeave its 2026 capex is $55-64B, so it adds up to $20-25B or less on its own.
Source ↗ - Metaplace 9
$20B-$25B of capex for $3B-$3.4B of revenue: the most leveraged pure-play bet on GPUs actually being rented.
Source ↗ - Grokplace 8
$20-25B guidance, rapid GPU-cloud scaling
Source ↗ - Z.aiplace 10
Guidance raised twice to $20-25 billion with $6.8 billion already gone in Q1 - a listed pure-play on GPU capacity that keeps selling out what it builds.
Source ↗ - Kimiplace 8
The smallest budget here ($20-25B) but arguably the highest chip purity — Q2 capex of $5.7B went to 'GPUs, GPU-related hardware and data centers.'
Source ↗
5 AIs put it here
02Worth knowing
What the AIs noticed while building the list.
Amazon's $20B raise in July was reported as mostly higher memory prices, so a bigger budget can mean dearer chips rather than more of them.
Investors did not reward bigger budgets this season: Microsoft cut its plan and rose, Alphabet held its plan and fell about 12%, and Meta raised its floor and fell about 7%.
Some chips never pass through the buyer's own spending line: Oracle has $75B of customer-prepaid or customer-supplied GPU contracts, and Meta holds about $62.2B of rented capacity from CoreWeave and Nebius.
Record budgets did not protect chip stocks: a Nov 5 press release has the SOX index down 12.25%, while Benzinga has the SOXX fund down 12.8%, its steepest day since 2020.
03Questions people ask
Who has the most AI chips?
Google, with the equivalent of about 5 million Nvidia H100s, roughly 25% of the world's total, mostly its own TPUs. Counting Nvidia chips alone, one tally puts Microsoft first at about 3.2 million.
Source ↗Who makes the most AI chips?
Nvidia designs the most, with 63% of AI computing power shipped in Q2 2026 on Epoch AI's count, ahead of Google at 25%.
Source ↗Who sells the most AI chips?
Nvidia, with $89.0B of data-centre revenue in the quarter to July 26, 2026. Its share is about 75% by revenue and 63% by computing power, which is why sources differ.
Source ↗Who manufactures AI chips?
Mostly TSMC, which took 72.5% of global foundry revenue in Q2 2026 and makes chips for Nvidia, AMD and Apple. Nvidia owns no factory of its own.
Source ↗04How we counted
In plain words, no formulas.
1The question
Top 10 AI Spenders Whose Budgets Really Reach Chipmakers, November 2026
2Why this topic
It separates headline AI budgets from the money likely to reach chipmakers and shows where companies surprised Wall Street, which no spending table does.
3Where it applies
Global listed AI and cloud companies
4What we looked at
- 2026 capital spending guidance in dollars
- Estimated share of that spending that goes to chips rather than buildings, power or land
- Gap between company guidance and analyst estimates before the latest earnings
- How the company's own stock moved after it raised spending
- Company earnings releases and filings — 2026 or fiscal-year capital spending guidance as each company states it
- CNBC report on Microsoft and Alphabet results — Latest guidance changes, and whether quarterly capital spending came in below analyst estimates
- CFO statements collected by Codema — Disclosed share of spending on servers, networking and short-lived assets such as CPUs and GPUs; a secondary compilation
- BNP Paribas AI capex cost chain — Split of each $100 of AI spending: $50 to semiconductor hardware, the rest to power, cooling, construction and networking
5Who ranked
The leading AIs, each building its own list of ten independently and naming a source for every place: Claude, Meta, Grok, Gemini, Nova, MiMo, Z.ai, Kimi, DeepSeek, Qwen.
6How the places were set
The same entry written differently by different AIs was counted as one. The more AIs put an entry on their list, and the higher they put it, the higher its place. "AI agreement" shows how many of the AIs had it on their own list. Sources are only those the AIs found themselves; nothing was added by hand.
05Each AI's own list
🇺🇸 Claude10 places · 3 notes
The same question, answered by Claude alone. The final list above combines all the AIs.
- 1Amazon (AWS)
The largest 2026 budget, about $220B, raised from $200B in July. Management says 2027 will be 'meaningfully more'.
Source ↗ - 2Microsoft
It cut 2026 capex to $170B from $175B and still rose after earnings. Its CFO says spending accelerates in H1 2027 because components cost more.
Source ↗ - 3Alphabet (Google)
It held a $195-205B range and the stock fell about 10-12%, the sharpest punishment of any spender this season.
Source ↗ - 4Meta Platforms
It raised only the floor of its range ($130B to $135B) and still lost 3% on the first reaction, then rose 3% a week later.
Source ↗ - 5Oracle
Fiscal 2027 gross capex of $90-95B, but its CFO says net cash capex is 'not more than $70B'. Remaining performance obligations are $664B.
- 6CoreWeave
Its $35-39B budget is small, but it is nearly all GPUs and the buildings are leased.
Source ↗ - 7Alibaba
The biggest Chinese spender: RMB190B of a RMB380B three-year plan is spent, and June-quarter capex rose 75%.
Source ↗ - 8SpaceX / xAI
It sits in the tracker of Q4 capex guided up to $270B for Big Tech and SpaceX combined, but I could not find a firm 2026 figure.
Source ↗ - 9Nebius
Combined with CoreWeave its 2026 capex is $55-64B, so it adds up to $20-25B or less on its own.
Source ↗ - 10Tencent and Baidu
China's cloud trio adds roughly $35B of spending beyond Alibaba, but no company-level guidance surfaced.
What Claude noticed
Only Meta raised its 2026 guidance this season, and only the floor. Microsoft cut and Alphabet held, which breaks the idea that every report brings a bigger budget.
Goldman puts 2027 AI capex at $1.2T against a consensus of about $1.1T, so the next gap is between the street and Goldman rather than between the street and the companies.
Investors punish negative free cash flow more than they reward a bigger budget: Alphabet fell 10-12% while Microsoft rose.
🇺🇸 Meta10 places · 3 notes
The same question, answered by Meta alone. The final list above combines all the AIs.
- 1Meta Platforms 2026
The only spender with signed chip deals bigger than most rivals' whole budgets: about $50B Nvidia plus $60B AMD.
Source ↗ - 2Amazon 2026
Largest headline budget at $220B, with a $578B AWS backlog that makes the chip bill look like a down payment.
Source ↗ - 3Microsoft 2026
The only big spender the market rewarded after earnings: Azure up 48% while capex came in below its own $50B-plus plan.
Source ↗ - 4Alphabet (Google) 2026
$195B-$205B guided, but 65% went to servers in Q3 and TPUs now ship in 78% of its AI servers.
Source ↗ - 5CoreWeave 2026
The purest GPU landlord: $35B-$39B of capex against only $12.4B-$13.2B of revenue guidance.
Source ↗ - 6Oracle 2026-2027
Gross $90B-$95B sounds huge, but net cash is capped at $70B: $20B-$25B is customer prepayments, not Oracle cash.
- 7Tencent 2026
Q2 alone was RMB 52.8B, up 176% year on year: at that pace Tencent is a $29B-plus annual spender without ever giving a formal full-year guide.
Source ↗ - 8Alibaba 2026
The smallest of the China giants on paper, but one June quarter alone was RMB 67.7B: 53.7% of its whole prior fiscal year.
Source ↗ - 9Nebius Group 2026
$20B-$25B of capex for $3B-$3.4B of revenue: the most leveraged pure-play bet on GPUs actually being rented.
Source ↗ - 10Tesla 2026
Cut its guide instead of raising it: $20B, down from $25B, while Q3 capex still doubled to $4.7B.
Source ↗
What Meta noticed
Microsoft was the outlier in late October: capex $45.2B came in below the $50.7B street pencil and the stock rose 3-4% after hours, while Alphabet fell 12.6% and Meta fell 7%.
Alphabet's chip story is custom, not Nvidia: TPUs are projected at nearly 78% of AI servers shipped to Google in 2026, so much of its $195B-$205B reaches Broadcom and TSMC, not merchant GPU sellers.
The SOX crashed 12.25% to 9,011.14 on Nov 5, its largest one-day fall since March 2020, even though the spenders had just reaffirmed higher budgets: chip stocks fell on policy and de-grossing, not on cancelled orders.
🇺🇸 Grok10 places · 4 notes
The same question, answered by Grok alone. The final list above combines all the AIs.
- 1Amazon 2026 CapEx
~$220B cash CapEx guide, Q3 alone $77B, still the largest absolute pipeline into GPUs and Trainium
Source ↗ - 2Alphabet 2026 CapEx
$195-205B range held, 65% of technical infrastructure already servers and networking
Source ↗ - 3Microsoft 2026 CapEx
~ $170B calendar guide, ~75% short-lived assets (CPUs/GPUs) — highest disclosed chip intensity
Source ↗ - 4Meta 2026 CapEx
$135-145B range, floor raised, heavy Llama/superintelligence server demand
Source ↗ - 5Oracle 2026/27 CapEx
$90-95B fiscal guidance, pure cloud/AI capacity build with large GPU deployments
- 6CoreWeave 2026 CapEx
$31-39B pure-play GPU cloud spend — almost all dollars flow straight to chips
Source ↗ - 7Tesla 2026 CapEx
$20B guide, Dojo and AI inference silicon plus data-center build
Source ↗ - 8Nebius 2026 CapEx
$20-25B guidance, rapid GPU-cloud scaling
Source ↗ - 9Micron 2026 CapEx
Memory maker but heavy HBM/AI capacity spend that enables the entire chip supply chain
Source ↗ - 10Broadcom / custom ASIC buyers (collective)
Custom silicon orders from hyperscalers form a large secondary channel that reaches TSMC and packaging houses
Source ↗
What Grok noticed
Four hyperscalers alone guide to roughly $720-740B of 2026 CapEx — more than many national infrastructure budgets — with the majority now short-lived compute rather than buildings.
Microsoft disclosed ~75% of its latest quarterly CapEx went to CPUs/GPUs; Alphabet ~65% to servers and networking — the clearest public splits showing money reaching chipmakers.
Despite record spend, free cash flow at Amazon, Alphabet and Meta turned negative, triggering equity and debt raises and stock-price punishment even as cloud growth accelerated.
Goldman and others see the broader hyperscaler AI CapEx trajectory continuing higher into 2027 ($1.1-1.2T range), implying the 2026 budgets are not a peak.
🇺🇸 Gemini10 places
The same question, answered by Gemini alone. The final list above combines all the AIs.
- 1Microsoft
Funneling an industry-leading $115 billion straight into silicon across Nvidia Blackwell clusters and custom Maia accelerators.
- 2Meta Platforms
Directs 68% of its budget into pure compute silicon, avoiding third-party commercial cloud overhead to train open-source Llama models.
- 3Alphabet
Splits its $110B hardware outlay evenly between commercial Nvidia GPUs and custom in-house TPU v5p/v6e silicon co-designed with Broadcom.
- 4Amazon
Boasts the world's largest $210B total capex envelope, though heavy civil real estate and logistics keep chip conversion closer to 51%.
- 5Oracle
Transforms 65% of its surging $55B budget directly into high-density GPU bare-metal clusters dedicated to leading frontier model builders.
- 6Tesla
Dedicates $12.4B straight to Nvidia hardware and custom AI chips for Cortex training superclusters and autonomous driving fleets.
- 7Alibaba
Ramps cloud capital spending by over 30% to feed proprietary Qwen models, deploying both international silicon and domestic Chinese accelerators.
- 8Tencent
Tripled quarterly infrastructure run-rates to secure high-bandwidth memory and training accelerators for its multimodal Hunyuan model.
- 9Baidu
Maintains a compact $9B budget with 64% routed into silicon, prioritizing proprietary Kunlunxin AI accelerators and Apollo autonomous computing.
- 10Apple
Capped facility spending at $9B while directing a staggering $45.8B into R&D and proprietary M-series silicon for Private Cloud Compute.
🇺🇸 Nova10 places · 3 notes
The same question, answered by Nova alone. The final list above combines all the AIs.
- 1Alphabet (GOOGL)
$185B total capex with $130B directly tied to AI chips – the clearest chip focus among giants
- 2Amazon (AMZN)
$220B capex where two-thirds ($145B) targets AI compute – logistics drag limits pure chip focus
Source ↗ - 3Microsoft (MSFT)
$170B capex with 66% ($112B) on GPUs/CPUs – Azure’s infrastructure play
Source ↗ - 4Meta (META)
$145B capex where 44% ($64B) funds custom AI silicon – partnership-driven acceleration
Source ↗ - 5Oracle (ORCL)
$58.8B capex with 40% ($23.5B) on cloud AI infrastructure – database-to-accelerator pivot
Source ↗ - 6Alibaba (BABA)
$126.9B capex but opaque chip split – domestic China dynamics limit transparency
- 7SpaceX (private)
$88B capex funneled through xAI – satellite-AI synergy creates unique exposure
Source ↗ - 8Tencent (00700.HK)
HK-listed $54.6B capex with growing AI cloud share – China’s silent infrastructure builder
Source ↗ - 9ByteDance (private)
Unlisted $40B+ capex fuels TikTok AI – regulatory risk clouds transparency
Source ↗ - 10IBM
$37.8B capex with 30% ($11.3B) on AI-powered enterprise infrastructure
Source ↗
What Nova noticed
Microsoft’s two-thirds chip allocation sets a de facto industry standard – expect 60-70% targets from peers
Private players (SpaceX, ByteDance) now control $130B+ of AI infrastructure spend – regulatory scrutiny to rise
Meta’s custom chip bets could redefine AI efficiency – MTIA may deliver 2x performance-per-dollar by 2027
🇨🇳 MiMo10 places · 4 notes
The same question, answered by MiMo alone. The final list above combines all the AIs.
- 1Intel stock (INTC, 2026)
Analysts' average target sits 62% above the price while their own headlines say 'urge caution on buying the dip'.
Source ↗ - 2AMD stock (2026)
Targets range from 31% to 160% upside on the same shares — and the stock fell after a blowout quarter.
Source ↗ - 3NVIDIA stock (NVDA, 2026)
54 analysts say Buy at an average target of $305 — the same week the chip index fell 12% in a day.
Source ↗ - 4Micron stock (MU, 2026)
Sank 17% in a single day of the crash — and no two sources agree on its consensus target.
Source ↗ - 5ASML stock (2026)
Morgan Stanley trimmed its target and still called the shares Overweight — the disagreement is inside one bank.
Source ↗ - 6TSMC stock (TSM, 2026)
A $265 billion campus announcement meets a $53.3 billion monthly sales hurdle — the biggest bet against the biggest bar.
Source ↗ - 7Broadcom stock (AVGO, 2026)
Jefferies once crowned it the top chip stock at $480 — then November's crash knocked 17% off a peer in a day.
Source ↗ - 8BE Semiconductor stock (Besi, 2026)
Morgan Stanley cut its target by 20% — from €81 to €65 — in one September move.
Source ↗ - 9Infineon stock (2026)
Downgraded to Equal Weight in the same breath that upgraded Synopsys — one bank, two directions, one sector.
Source ↗ - 10KLA stock (KLAC, 2026)
Made a 'best stocks' list on Nov 2 data — three days before the sector's worst day since 2020.
What MiMo noticed
The whole disagreement table rests on one number: SOXX fell 12.8% on Nov 5 — its steepest day since 2020 — with 28 of 30 chip stocks in bear markets, so nearly every 'average target' is now stale by a week.
This is the second shock of 2026 for chips: July was already the sector's worst month since 2008 (SOXX −18.6% month-to-date), and BofA told clients to buy the reset — the same call, twice.
The contrarians were early and famous: Michael Burry bought SOXX puts in April 2026 against a historic chip rally — the snap he bet on arrived seven months later.
Intel is the widest disagreement in the sector: the same analysts whose average target implies 62% upside were telling clients to be careful buying the dip.
🇨🇳 Z.ai10 places · 3 notes
The same question, answered by Z.ai alone. The final list above combines all the AIs.
- 1Amazon (AWS) - AI capex 2026
The only spender above $200 billion - and it raised the budget by $20 billion just to cover memory-price inflation, with a $496 billion AWS backlog still outrunning capacity into 2027.
Source ↗ - 2Microsoft (Azure) - AI capex 2026
The only giant that trimmed its 2026 budget - to $170 billion - because third-party GPUs and CPUs arrived late, not because it wanted to spend less; fiscal 2027 is guided to ~$255-260 billion.
Source ↗ - 3Alphabet (Google Cloud) - AI capex 2026
Guidance went from $175-185 billion to $195-205 billion in five months - and because its TPU pipeline runs through Broadcom, part of every dollar reaches a second chipmaker, not just Nvidia.
- 4Oracle (OCI) - AI capex FY2026-27
A $638 billion backlog - about 11 times its $55.7 billion FY26 capex - means nearly every dollar is contracted with a customer before it ever reaches a chipmaker.
Source ↗ - 5Meta Platforms - AI capex 2026
$135-145 billion and its first negative free-cash-flow quarter (-$2.12 billion) - the superintelligence bet now costs more than the core business generates.
Source ↗ - 6SoftBank (Stargate / OpenAI) - 2026
Its $500 billion Stargate envelope has about $400 billion of committed three-year spend across seven US sites plus the UAE - the biggest chip budget that sits on nobody's earnings call.
Source ↗ - 7CoreWeave - AI neocloud capex 2026
Capex of $35-39 billion against guided revenue of $12.4-13.2 billion - the purest GPU-per-dollar ratio in the group, with a ~$104 billion backlog behind it.
Source ↗ - 8Alibaba Cloud - AI capex 2026 (China)
Half of its RMB 380 billion three-year plan is already spent; June-quarter capex rose 75% while net income fell 75% - and the stock is down 25% year to date anyway.
Source ↗ - 9Tencent - AI capex 2026 (China)
Operating capex jumped 190% in a single quarter to RMB 51.8 billion - and with US chip curbs in force, a growing slice flows to domestic silicon instead of Nvidia.
- 10Nebius - AI neocloud capex 2026
Guidance raised twice to $20-25 billion with $6.8 billion already gone in Q1 - a listed pure-play on GPU capacity that keeps selling out what it builds.
Source ↗
What Z.ai noticed
Chips, not buildings, are the bill: GPUs alone are about 56% of the $37.2 billion cost of a 1-gigawatt AI data center, and PwC puts ICT equipment at ~70% of all data-center capex in 2026 - every $1 of construction commits roughly $12 of future chip-and-server spend. That is why this ranking weighs chip conversion, not raw dollars alone.
The market's verdict flipped twice this year: Amazon lost about 10% for guiding $200 billion in February, Microsoft's shares rose after trimming 2026 capex to $170 billion in October, and Alphabet sank 10% after hours for holding guidance - investors now punish both over-spending and under-delivery.
The largest single chip buyer is not even listed: OpenAI plans about $50 billion of compute in 2026, triple last year - which is why SoftBank's Stargate envelope, not its own balance sheet, is the number to watch, and why Microsoft is re-designing its own Maia silicon to blunt the bill.
🇨🇳 Kimi10 places · 4 notes
The same question, answered by Kimi alone. The final list above combines all the AIs.
- 1Microsoft (2026)
Guides ~$255-260B for FY2027 and says two-thirds of every capex dollar already goes to 'short-lived' assets — mostly CPUs and GPUs.
Source ↗ - 2Amazon (2026)
Reiterated ~$220B for 2026 with 2027 'meaningfully more' — and AWS growing 39%, its fastest in five years, means the chips are spoken for.
Source ↗ - 3Alphabet / Google (2026)
Doubled its capex plan in 18 months to $195-205B — and even selling its own TPUs to cloud customers now counts as chip demand.
Source ↗ - 4Meta Platforms (2026)
Raised its capex floor four quarters in a row to $135-145B and is still 'supply constrained on compute' — so it rents even more chips from neoclouds on top.
Source ↗ - 5Oracle (FY2027)
Guides $90-95B for FY2027 — but $75B of its AI contracts are customer-prepaid or customer-supplied GPUs, so the chips arrive even when Oracle's cash doesn't pay for them.
Source ↗ - 6Alibaba (2026)
Halfway through its RMB 380B ($56B) three-year AI plan after 17 months — and already says it will exceed it, while profit fell 99.7% to pay for it.
Source ↗ - 7CoreWeave (2026)
The purest chip buyer on the list: ~$37B of capex that is almost entirely GPUs, with a $104B backlog pre-paying for them.
Source ↗ - 8Nebius Group (2026)
The smallest budget here ($20-25B) but arguably the highest chip purity — Q2 capex of $5.7B went to 'GPUs, GPU-related hardware and data centers.'
Source ↗ - 9Tesla (2026)
The only top-10 name that CUT — $20B, down from 'more than $25B' — and much of it buys factories and robotaxis, not chips.
Source ↗ - 10Apple (2026)
The $3T company that barely builds: the 'biggest wildcard' among the big five — it rents AI compute rather than guiding a mega-budget.
What Kimi noticed
The chip-vs-concrete split is the least agreed number in AI: PwC says ICT equipment is ~70% of 2026 data-center capex, BNP Paribas says chips+servers are 50%, and a Goldman-anchored analysis argues only ~25% reaches chips — the difference between ~$550B and ~$190B of annual chip demand.
Microsoft made its capex look smaller without spending a cent less: extending useful lives from 15 to 25 years shifts leases out of the capex line, cutting reported 2026 capex from ~$190B to ~$175B while two-thirds of the real budget still buys CPUs and GPUs.
Some of the biggest 'chip budgets' never touch the spender's cash: Oracle's customers prepaid or supplied $75B of GPUs, and Meta holds ~$62.2B of neocloud contracts — chips bought by CoreWeave and Nebius instead.
The market now rewards restraint and punishes raises: Microsoft jumped ~15% for holding capex flat in July while Meta fell ~16% on the same day for lifting its floor — yet total 2026 guidance still climbed ~77% to ~$725B for the big four.
🇨🇳 DeepSeek10 places · 4 notes
The same question, answered by DeepSeek alone. The final list above combines all the AIs.
- 1Alphabet (Google) 2026 AI capex
The $205 billion budget is the largest single chip-directed pool on this list: about 60% goes to servers, roughly $123 billion, before a single dollar of buildings or power.
Source ↗ - 2Amazon 2026 AI capex
The $220 billion budget grew by $20 billion in July for one reason — memory prices — the cleanest proof on this list that the chip line, not the building line, is what moves.
Source ↗ - 3Microsoft 2026 AI capex
Wall Street wanted $154.6 billion; Microsoft guided $190 billion, and $25 billion of the difference is memory and component costs — a capex beat made of chips.
Source ↗ - 4Oracle fiscal 2027 AI capex
A $95 billion reported capex budget against roughly $67.7 billion of analyst expectations, with more than 60% aimed at Nvidia GPU clusters — and the stock paid for it with a double-digit drop.
Source ↗ - 5CoreWeave 2026 AI capex
It guides $35–$39 billion of capex on $12.4–$13.2 billion of revenue — a capex-to-revenue ratio above 3x, the highest purity of chip spending on this list.
Source ↗ - 6Meta Platforms 2026 AI capex
Meta's $145 billion budget is the one where chip suppliers are literally buying the equity — Nvidia, AMD and Broadcom offered up to $5 billion — because the building and power bill is so large.
Source ↗ - 7Tesla 2026 AI capex
A $25 billion capex year with close to $20 billion tagged for AI — but the stock fell 14.5%, the sharpest rejection of any budget on this list.
Source ↗ - 8Alibaba 2026 AI capex
CLSA sees RMB220 billion, about $33 billion, of 2026 capex — China's largest listed AI capex line that visibly reaches chipmakers, even as the ADR stays under pressure.
Source ↗ - 9xAI (SpaceX) 2026 AI capex
xAI alone spent $23.5 billion in 2026, 82.7% of SpaceX's entire capex, and its Q2 AI capex more than doubled to $15.8 billion — a GPU landlord in all but name.
Source ↗ - 10Tencent 2026 AI capex
Tencent's AI capex more than doubled to RMB96.3 billion in H1 2026, and it is borrowing $5 billion mainly to buy chips — yet the stock hit a one-year low.
Source ↗
What DeepSeek noticed
Amazon added $20 billion to its 2026 capex in July and attributed the entire increase to higher memory costs, not to a bigger building plan — the single cleanest example of AI capex inflation being chip inflation.
Alphabet spends about 60 cents of every infrastructure dollar on servers, which at the midpoint of its $205 billion guide is roughly $120 billion of server purchases — before land, power or shell construction.
Meta's chip suppliers Nvidia, AMD and Broadcom offered to buy up to $5 billion of Meta equity while Meta was raising $95 billion to fund AI capex — a circular financing structure that ties the customer's budget directly to its suppliers' balance sheets.
CoreWeave's 2026 capex is more than three times its revenue, but its $104.2 billion revenue backlog means much of the GPU capacity is pre-sold before the chips are even installed.
🇨🇳 Qwen10 places · 3 notes
The same question, answered by Qwen alone. The final list above combines all the AIs.
- 1CoreWeave 2026
A pure GPU proxy with a $104B backlog that converts debt almost entirely into Nvidia orders.
Source ↗ - 2Anthropic 2026
$63 billion in bespoke financing arranged specifically to secure compute silicon rather than pour concrete.
Source ↗ - 3xAI 2026
Deploying $15.8B in a single quarter to build the world's largest liquid-cooled GPU clusters.
Source ↗ - 4Meta Platforms 2026
Back on offense with a $145B ceiling that prioritizes training Llama over traditional ad infrastructure.
Source ↗ - 5Amazon 2026
Raised its budget by $20B mostly to absorb soaring memory prices for custom Trainium and Nvidia racks.
Source ↗ - 6Alphabet 2026
Guiding $205B but taking a 10% stock hit as investors fear the cash is leaking into power plants instead of TPUs.
Source ↗ - 7Microsoft 2026
Cut its forecast to $170B as supply chain delays force it to extend the useful life of older servers.
Source ↗ - 8Oracle 2026
Surging 162% to build cloud regions, but weighed down by $130B in debt to finance the concrete.
- 9Tesla 2026
Slashing its budget to $20B to focus on Dojo and factories rather than standard cloud AI chips.
Source ↗ - 10Apple 2026
Outsourcing the AI race to partners with a $14B budget that barely registers against the hyperscalers.
Source ↗
What Qwen noticed
Goldman Sachs estimates that hyperscalers will spend 100% of their cash flow from operations on capital expenditures by the end of 2026.
Wall Street banks including Evercore and Bank of America now forecast that global AI-related capital expenditure will cross the $1 trillion threshold in 2027.
Amazon raised its 2026 capital spending to roughly $220 billion primarily to absorb soaring memory costs rather than just buying more GPUs.