Chips

Nvidia's $279 billion backs an estimated 37% of the world's 2027 AI memory

Morgan Stanley estimates put Nvidia, Alphabet and AMD at 85% of 2027 high-bandwidth memory, Nvidia alone at roughly 37%, backed by $279 billion in disclosed supply commitments. The DDR5 market is paying the toll now, and the next reading lands at Nvidia's November earnings.

Vincent JiangVincent Jiangthree-minute read
Share
Nvidia chief executive Jensen Huang speaking at CES, gesturing with both hands against an Nvidia and CES branded backdrop.
1 / 6Slide 1 of 6
Jensen Huang, Nvidia's chief executive, whose promise to double chip sales next year sits behind the $279 billion memory commitment.

The quarter Nvidia bought the memory market

Nvidia's−3.21% — Nvidia, down 3.21 percent today supply commitments, the purchases it has promised its suppliers, jumped from $119 billion to $279 billion in the quarter ended 26 July, an increase the company tied "primarily" to memory procurement 1. About $267 billion of it comes due by the end of fiscal 2029 1.

Jensen Huang said on 17 September that Nvidia will sell twice as many chips next year as this year 2. Twice the chips means twice the memory, and the memory is already spoken for.

Three buyers, 85 percent of the shelf

Procurement data compiled on 25 September pegs Nvidia's claims alone at roughly 37% of global 2027 high-bandwidth memory output 3. Morgan Stanley's+0.55% — Morgan Stanley, up 0.55 percent today split puts Nvidia at 37.3%, Alphabet−1.54% — Alphabet, down 1.54 percent today at 36% and AMD−10.90% — AMD, down 10.90 percent today at 12.1%, a combined 85% 45. Three companies hold the deeds to AI's scarcest input; everyone else farms the queue.

Three buyers hold 85% of 2027 HBM capacity; everyone else shares 15%

  • Estimate
0%10%20%30%40%Nvidia37.3%Alphabet36%Every other builder14.6%the queue behind three buyersAMD12.1%
Data
Value
Nvidia (estimate)37.3%
Alphabet (estimate)36%
Every other builder (estimate)14.6%
AMD (estimate)12.1%
Percent of projected 2027 global high-bandwidth memory capacity by buyer. All figures are Morgan Stanley estimates from channel checks, not company disclosures; 'Every other builder' is the residual.4,5

Every HBM wafer is a DDR5 wafer that is not

HBM eats upwards of three times the wafer capacity of ordinary DRAM, on scanners that cannot go back to server memory 53. Samsung has handed more than half its 4nm capacity to HBM4 base-die production, the maker side of this squeeze covered on 22 September 6. Samsung, SK hynix−5.27% — SK hynix, down 5.27 percent today and Micron−4.50% — Micron, down 4.50 percent today run more than 90% of the world's DRAM wafers 3, their 2027 HBM books are full, and new capacity lands in late 2027 4.

The toll is arriving downstream. Tier-one server builders have cut inventories to eight to 11 weeks of supply 3, with PC, server and car buyers queued behind them. German retail DDR5 sits 544% above its July 2025 baseline, and 32GB kits that bottomed near $72 in the US now run $409 7.

The other camp calls the shortage a sales pitch

Acer−0.34% — Acer, down 0.34 percent today chief executive Jason Chen says memory makers are exaggerating constraints to protect AI-inflated margins, and expects prices to fall after mid-2027 as China's CXMT ramps 8. His own forecast concedes PC prices climb 5 to 20% through the fourth quarter first 8. SK hynix, holding an estimated 60 to 70% of Vera Rubin's HBM4 volume under a June partnership reported at $500 billion 95, has little incentive to agree.

November is the checkpoint

Operating cash flow more than halved last quarter, from $50.3 billion to $24.1 billion, as the prepayments left the building 10. Gross margin is guided to a 71 to 72% trough in the fiscal fourth quarter before price increases bite 1. Nvidia flags the rising obligations as a risk if data-center deployments slow 3. The queue for the last 15% forms before the late-November earnings call restates the number, and Vera Rubin volume shipments start in the first half of 2027 3.

Operating cash flow halved last quarter as the memory prepayments left

$0B$20B$40B$60BQ4 '23Q2 '24Q4 '24Q2 '25Q4 '25Q2 '26$24.08B
Data
Operating cash flow
Q3 '23$7.33B
Q4 '23$11.5B
Q1 '24$15.35B
Q2 '24$14.49B
Q3 '24$17.63B
Q4 '24$16.63B
Q1 '25$27.41B
Q2 '25$15.37B
Q3 '25$23.75B
Q4 '25$36.19B
Q1 '26$50.34B
Q2 '26$24.08B
Nvidia operating cash flow by fiscal quarter, in US dollars. The fiscal second quarter of 2026 ended 26 July 2026; the drop to $24.1 billion from $50.3 billion reflects the prepayments behind the supply commitments.11

Get Chips by email, free

How AI demand is reshaping chipmakers, memory and the fabs behind them. In our daily email, every weekday.

Loading NVDA

More about NVIDIA

NVDAFiscal Q2 2027
Fiscal Q2 2027 · quarter to 26 Jul 2026 · changes vs a year earlier

Revenue rose 106% to $96.2B, 92.5% of it from Data Center, “driven by the ramp of our Blackwell Ultra infrastructure”. Gains on equity stakes of $7.8B lifted net income to $59.7B.

Revenue$96.2B▲ 106%
Gross margin75.0%▲ 2.6 pts
Operating margin66.2%▲ 5.4 pts
Net income$59.7B▲ 126%
Hyperscale$48.7B▲ 102%AI clouds & enterprise$40.3B▲ 138%Data Center$89.0B▲ 117%Edge Computing$7.2B▲ 27%Revenue$96.2B▲ 106%Gross profit$72.1B75.0% marginCost of revenue$24.1BOther income$7.8Bequity-stake gainsOperating income$63.7B66.2% marginOperating expenses$8.4BNet income$59.7B▲ 126%Tax$11.8B16.5% rateR&D$7.1BSG&A$1.4B
Show as a table
LineValue
Revenue$96.2B
Gross margin75.0%
Operating margin66.2%
Net income$59.7B
Hyperscale$48.7B
AI clouds & enterprise$40.3B
Data Center$89.0B
Edge Computing$7.2B
Gross profit$72.1B
Cost of revenue$24.1B
Other income$7.8B
Operating income$63.7B
Operating expenses$8.4B
Tax$11.8B
R&D$7.1B
SG&A$1.4B

Reader comments

0 comments

    Chips

    Fractile, valued at $7 billion, says its product chip goes to the factory in a few months

    Founder Walter Goodwin told the 20VC podcast that test chips are back and trial runs continue, with first systems promised for next year. He said most of the $720 million round is for locking up parts years ahead, from corners of the supply chain his rivals have left.

    Chips

    MaxLinear launches a power chip to save up to half a watt per AI optical module, its answer to Marvell

    The chip is genuine engineering on a thin revenue line: it saves up to half a watt per module, and its job is to help sell MaxLinear's pricier Rushmore DSP into the same socket, the company's first product answer to Marvell's $115 billion Lumentum deal. Samples in February 2027 are the first date the strategy can show up in hardware.

    Chips

    Samsung Electro-Mechanics is 62% off its June peak even as capacitor waits hit 61 weeks

    Korea's press ties the slide to fears of tighter US data-center rules after the midterm elections, but the order book runs the other way: 61-week waits for its capacitors and Nvidia money behind its new substrate lines. The January restock date, and whether about 50 long-term supply contracts hold, will show which side is right.

    Chips

    Bessent gives Nvidia's board 10 days to answer whether suspect China sales inflated its earnings

    Treasury Secretary Scott Bessent has given Nvidia's board 10 days to answer whether suspect China sales inflated its earnings, two days before the first results since Jensen Huang resigned. What moves the stock on Tuesday is the Singapore-billed revenue Nvidia no longer breaks out, presented by interim CEO Colette Kress, who signed the deals under investigation as CFO.

    Chips

    Nvidia director Mark Stevens sold $946 million of stock with no trading plan before the Huang scandal

    He sold near $220 in September; the stock trades near $163, the dip buyers in between are carrying the difference, and Nvidia's first report under interim CEO Colette Kress lands Nov. 17.

    Chips

    Jim Cramer's CNBC club will sell Intel and buy AMD as Intel guides $15-16 billion, below $17 billion expected

    The call came a day after Mercury Research put Intel's desktop CPU share at 70.1%, a gain the research firm says may only be Intel refilling supply in a shrinking PC market. The money is moving to the server CPUs that run AI agents, and AMD is guided to outsell Intel this quarter for the first time.

    Subscribe to TheInference for free