J.P. Morgan’s Michael Cembalest wrote in the Eye on the Market note dated 14 November that Nvidia “has only a 39% share of future frontier lab compute agreements compared to its 80%+ market share from 2023-2025”. The note’s chart, sourced to J.P. Morgan Asset Management, gives Broadcom custom chips 24%, Google TPUs 16%, AMD GPUs 15% and Amazon Trainium 6%.
Added together, custom chips from Broadcom, Google and Amazon hold 46% of those agreements, 7 points more than Nvidia’s 39%, by our arithmetic. The note is Cembalest’s own view; its disclaimer says it “in no way constitutes J.P. Morgan Research”.
Business Pill · Switching costs explained
An evergreen Business Engineer explainer on switching costs, why it is hard to leave an AI supplier. It teaches the concept only and does not discuss Nvidia, J.P. Morgan’s note or any company in this piece.
The key insight: As we read the chart, the larger shift is to custom silicon rather than to AMD: Broadcom, Google and Amazon chips together hold 46% of frontier labs’ future agreements, against 15% for AMD GPUs and 39% for Nvidia, by our arithmetic.
What the Note Shows
The chart is titled “Frontier Lab future compute agreements” and is measured in gigawatts. Its single bar reads at roughly 41 GW, so Nvidia’s 39% is about 16 GW, by our arithmetic on a chart-read total that is approximate.
Everything other than Nvidia comes to 61% of the bar, and GPUs from Nvidia and AMD together come to 54%, by our arithmetic. Nvidia’s 39% is less than half of the 80%+ market share the note gives it for 2023 to 2025.
The note also says Nvidia “earned 83% of AI chip revenues from 2023-2025 with gross margins of 75% and operating margins of 66%”. A box beside the chart adds: “Of the seven large users and purchasers of accelerators (the five hyperscalers and the two frontier labs), only SpaceX is not trying to build their own”.

What Was Posted on X
Coin Bureau posted a summary of the note at 07:45 UTC on 15 November, describing it as a JPMorgan note: the 39% share, the other chip shares, the 3-day AMD onboarding account, the SpaceX line, the 58% and $214 billion to $422 billion figures and the Baird and Evercore target cuts. We checked each figure against the note itself and Nvidia’s 10-Qs; the note’s own label for the $214 billion and $422 billion is off-balance sheet long term debt and lease liabilities, and the 10-Qs list those items as commitments.
Puckrin posted page 4 of the note as an image at 07:54 UTC on 15 November, with the 39% paragraph, the stacked bar and the 3-day passage; the image matches the PDF we read. The comments on moats and switching costs in the post are the poster’s own reading, not the note’s wording.
Why the Note Says Labs Are Adding Suppliers
Cembalest calls this “the most important section of this note”. The note cites an Epoch AI inference index that “shows AMD at the top” on modelled performance per chip, and says Nvidia’s Rubin Ultra series “may change that when rolled out in 2027”.
The note relates what it “heard from one of the frontier labs last week”: “a year ago it would have taken 6-12 months to onboard and start running models on AMD GPUs”. It goes on: “Now, using their latest AI model to do the engineering, they can onboard new models on AMD GPUs in just 3 days”. The lab is not named.
The note sets this against Nvidia’s case that “its superior tokens per watt and superior software ecosystem (CUDA) made switching uneconomic”, and says AMD’s MI455X results “may have eroded at least part of that advantage”.
What Nvidia’s 10-Q Shows
The note says Nvidia’s “off-balance sheet long term debt and lease liabilities rose from $214 bn to $422 bn” from Q1 to Q2, and to $530 billion with land, power and shell guarantees. Nvidia’s two latest 10-Qs label these items commitments and guarantees, not debt.
The fiscal Q1 2027 10-Q, for the quarter to 26 April, lists supply and capacity commitments of $119 billion, cloud services of $30 billion, other vendor commitments of $6 billion, investment commitments of $27 billion and $32.4 billion of leases not yet commenced. Those sum to $214.4 billion, by our arithmetic.
The fiscal Q2 10-Q, for the quarter to 26 July, has a commitments table totalling $366 billion and an Additional Commitments table of $56 billion, which sum to $422 billion, by our arithmetic. Adding the $108.5 billion maximum gross exposure on its guarantees gives $530.5 billion.
The total roughly doubled, up $207.6 billion, and supply and capacity, from $119 billion to $279 billion, account for 77% of that rise, by our arithmetic. The filing says some supply agreements “may be cancelable, rescheduled, or adjustable” before firm orders. Our piece on Nvidia’s supply commitments has the full breakdown.
Of the guarantees, $105 billion is a cap on credit support entered in August 2026 for an OpenAI affiliate’s leases at SB Energy’s Ohio campus, the filing says. The note’s other figure, that Nvidia spent 58% of operating cash flow on equity investments and M&A over three quarters, cites Bloomberg and company filings; we did not recompute it.
What Brokers Have Done
Baird cut its Nvidia target to $200 from $500, a 60% cut, Investing.com reported on 11 November. Evercore ISI lowered its target to $186, Investing.com reported on 12 November; the note’s table, sourced to Bloomberg, gives Evercore’s 1 July target as $413, a 55% cut.
Cembalest still writes: “I agree with the notion that Nvidia looks interesting for long term holders at 32% below its all-time high and with a PEG ratio of just 0.2x”. Our tracker of broker target cuts lists the other moves.
The Structural Read
As we read the 10-Qs, most of the note’s jump from $214 billion to $422 billion is supply: commitments rose from $119 billion to $279 billion, 77% of the $207.6 billion rise, by our arithmetic.
As we read the onboarding account, it is one unnamed lab’s switching time on AMD GPUs, from 6-12 months to 3 days, which the note pairs with an Epoch AI inference index; it is not a measured industry figure.
Michael Cembalest, Eye on the Market, J.P. Morgan, 14 November 2026
“Based on the bar chart above, the economics (i.e., total cost of ownership normalized to output per GW) may be shifting as frontier labs vote with their feet.”
Three Implications
FOR NVIDIA HOLDERS The note gives Nvidia 39% of future frontier-lab agreements, against 80%+ market share in 2023 to 2025, and a PEG ratio of 0.2x.
FOR CUSTOM-CHIP MAKERS Broadcom custom, Google TPUs and Amazon Trainium together hold 46% of the chart, by our arithmetic, with Broadcom custom the largest at 24%.
FOR BALANCE-SHEET READERS Nvidia’s fiscal Q2 10-Q shows $366 billion of commitments, $56 billion of additional commitments and $108.5 billion of maximum guarantee exposure.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework says: “Three competitive categories within one layer: generalist substrate (NVIDIA), hyperscaler custom (TPU, Trainium, MTIA), merchant custom challengers (Cerebras, Groq, SambaNova).”
As we read it, the note’s chart fills the first two categories: Nvidia GPUs at 39% and hyperscaler TPUs and Trainium at 22%, by our arithmetic, with Broadcom custom chips a further 24%; the framework does not discuss this note.
What Is Not Established
The pages we read do not name the frontier labs or list the agreements behind the 39%, and they do not give the dates those agreements run to. The 3-day AMD onboarding figure comes from one unnamed lab, as the note relates it; we could not verify it.
The $214 billion and $422 billion match sums of the filings’ commitment tables, by our arithmetic, but the note does not show its own calculation. J.P. Morgan and Nvidia were not contacted for this piece. Our Nvidia hub tracks the rest of the story.
The Bottom Line
J.P. Morgan’s Eye on the Market gives Nvidia 39% of frontier labs’ future compute agreements, against 80%+ market share in 2023 to 2025, and custom chips from Broadcom, Google and Amazon 46%, by our arithmetic. Nvidia’s own 10-Qs show commitments of $422 billion at 26 July, by our sum, up from $214.4 billion.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The 39% share, the chart shares, the frontier-lab onboarding account and the $214 billion, $422 billion and $530 billion figures come from J.P. Morgan’s Eye on the Market note by Michael Cembalest, dated 14 November 2026, read in full as a public PDF; the note says it is not J.P. Morgan Research. Commitment and guarantee figures come from NVIDIA’s 10-Qs for the quarters ended 26 April and 26 July 2026 on SEC EDGAR, the latest on file.
Broker targets come from Investing.com and the note’s table. Posts on X that circulated the note are shown as chatter, not as sources. Sums, shares and differences are our arithmetic. J.P. Morgan and Nvidia were not contacted for this piece.
Sources: J.P. Morgan, Eye on the Market: The Icarus Files (14 Nov 2026) · NVIDIA Form 10-Q, quarter ended July 26, 2026 (SEC EDGAR) · NVIDIA Form 10-Q, quarter ended April 26, 2026 (SEC EDGAR) · Investing.com: Baird downgrades Nvidia (11 Nov) · Investing.com: Evercore ISI cuts Nvidia (12 Nov)









