Nvidia Faces “Too Many Semiconductors and Not Enough Data Center Shells” by 2028, Goldman Warns Roxanna Maglangit Tue, November 10, 2026 at 11:35 AM EST 5 min read NVDA -6.55% GS -0.81% SMCI -4.20% Explore stocks on Coinbase Trading disclosure Trading disclosure The above button links to Coinbase. Yahoo Finance is not a broker-dealer or investment adviser and does not offer securities or cryptocurrencies for sale or facilitate trading. Coinbase pays us for certain activity generated through this link. Prices displayed are informational.
Goldman Sachs warned that by 2028 there won't be enough data center shells for all the AI chips being made, and Super Micro Computer's CFO says shells already set the pace of its deliveries.
For NVIDIA, chips waiting on buildings would hit inventory and pricing long before any customer cancels an order.
NVIDIA's inventory more than doubled to $31.6 billion over the year to July 2026, about in step with revenue, and gross margin was 75% in the latest quarter.
NVIDIA's Nov. 17 report is the next test: inventory against revenue, and the margin guide for the January quarter.
NVIDIA (NVDA:NASDAQ) had lost nearly $1.4 trillion of market value from its October record by the close on Nov. 5, after a 9% drop that day on midterm data center fears.
Now Goldman Sachs sees a problem that could hurt NVIDIA even if demand holds up.
In a note read out on CNBC's Squawk on the Street on Nov. 6, Goldman pointed to "the view that by 28 we won't have enough data center shells, that is land, power, shell and therefore will likely have too many semiconductors and not enough data center shells for them to go into."
That's a bear case even a demand bull has to take seriously.
TIKR tracks 62 analysts on Nvidia. How many say "Buy"?
See the full breakdown of Buy, Hold, and Sell ratings on Nvidia, along with Wall Street's revenue and earnings estimates, all in one place. More than 500,000 investors use TIKR to research stocks like this one. See Nvidia's Analyst Ratings
On its fiscal first-quarter call on Nov. 3, Super Micro Computer (SMCI:NASDAQ) raised its full-year revenue outlook to $72 billion to $82 billion, from $65 billion to $72 billion. But CFO Mark Long was clear about what sets the pace: "The binding constraints on AI build-outs across this industry right now are land, power and data center shells."
He said the range depends on "the pace at which customer facilities become ready to receive what we've already manufactured." And about $2.3 billion of the quarter's revenue was catch-up from earlier deployments, so Super Micro's results already move with its customers' buildings.
Elon Musk's list of bottlenecks is longer still. Replying on X to a post that summed up his Dwarkesh Patel interview as saying the industry would soon be "making more AI chips than we can actually turn on," Musk added: "And electrical equipment, cooling, networking and batteries."
That matters for NVIDIA, since analysts expect its revenue to keep climbing: $412 billion in the year ending January 2027, $692 billion in fiscal 2028 and $903 billion in fiscal 2029, which covers most of Goldman's 2028:
I wouldn't expect analysts to cut the next two years' estimates on Goldman's note, since NVIDIA says it can't make enough chips to meet demand through fiscal 2028. Fiscal 2029 is the year to watch, and only 38 analysts forecast it.
If buyers' buildings aren't ready, I'd expect them to push deliveries back, leaving chips on NVIDIA's shelves. Its inventory has more than quadrupled in two years:
That looks alarming until you set it against sales. Revenue more than doubled from a year earlier to $96.2 billion in the quarter to July 26, 2026, and inventory grew about as fast, to $31.6 billion from $15.0 billion.
What's coming in worries me more: NVIDIA's latest quarterly report shows $279 billion of supply and capacity commitments, mostly memory and manufacturing, up from $119 billion a quarter earlier. It's on the hook for that supply whether or not its customers have somewhere to put it.
A pile of chips its customers can't use would leave NVIDIA two choices: cut prices or write the inventory down. Either one hits gross margin, which has held between 73% and 75% for the past four quarters:
The one big dip, to 60.5% in the quarter to April 2025, is the closest thing to a preview. NVIDIA took a $4.5 billion charge for excess H20 inventory and purchase obligations after the US required a license to sell those chips in China.
The next few quarters will be harder to read, since management already expects gross margin to bottom at 71% to 72% in the January 2027 quarter as memory prices climb.
Demand still runs well ahead of supply: NVIDIA cites a cloud backlog above $2 trillion, and in May, management said H100 rental prices were up 20% for the year. But rising rents are what a shell shortage looks like, since a plugged-in chip earns more when powered space is scarce. Rents are only paid on chips that already have a home, and Goldman's warning is about the ones still waiting for one.
Goldman's warning is the right one to worry about, since stranded chips would hit NVIDIA's inventory and pricing long before any customer cancels an order. But NVIDIA's own numbers don't show a pile-up yet, so I wouldn't sell the stock on Goldman's note alone.
NVIDIA reports on Nov. 17. If inventory grows faster than revenue, or the January-quarter margin guide comes in below 71%, that's the first sign Goldman is right.
TIKR lets you forecast the future price of any stock in less than a minute. Just enter a few assumptions into TIKR's valuation model and see what NVIDIA could be worth. Start from Wall Street consensus estimates, or adjust the inputs to reflect your own view of the business. It's free to use.
