Published: August 27, 2026 | Reading Time: ~9 minutes | Channel: technology
Nvidia reported $96.2 billion in quarterly revenue — up 106% from a year ago — and the market's first instinct was to buy more. But the real story isn't the number. It's what Jensen Huang did next: he quietly agreed to pay $12.9 billion for Hugging Face, an open-source hub doing roughly $150 million a year, and then stood on camera and admitted gross margins are coming down from 75% to 72–73%. Read those two moves together and the narrative flips. The most dominant company in tech isn't reaching for more growth. It's building a wall.
That's not a bearish call on Nvidia. It's the opposite — it's the most honest admission from the C-suite that the free lunch is ending, and the company with the deepest pockets is buying the only thing that keeps it on top when it does: distribution, developers, and a reason for the market to stay dependent on its chips.
Let's get the facts straight first, because half the coverage has been sloppy.
On Wednesday night, The Information reported — citing a person familiar with the matter — that Nvidia had agreed to buy Hugging Face for $12.9 billion.¹ Business Insider separately reported the two sides were "in talks" and that a signed agreement hadn't yet materialized.² As of this writing, neither Nvidia nor Hugging Face has confirmed on the record. So let's be precise: this is a reported deal, not a done one. It could still atomize.
But assume it lands. What does Nvidia get for $12.9 billion?
Hugging Face is the closest thing the AI world has to a public utility for open-source models. Founded in 2016, it's where developers go to download, share, and deploy models — the anti-OpenAI. It raised $235 million in 2023 at a $4.5 billion valuation, with money from Salesforce Ventures, Alphabet's GV, and Nvidia itself.¹ It was generating about $150 million in annual revenue recently, up from roughly $100 million just two months earlier.¹
Do the math. $12.9 billion for $150 million in run-rate revenue is roughly an 86x price-to-sales multiple. That is not a growth acquisition. That is not a synergy play. That is a king writing a check to protect a crown that its own customers are trying to yank off its head.
Here's the part the bulls gloss over: OpenAI, Google, Amazon, and Anthropic are all building their own AI chips to reduce their dependence on Nvidia.³ Those four are Nvidia's largest customers. They are simultaneously its biggest source of revenue and its most motivated competitors. When your best customers start vertically integrating to stop paying you, you have exactly two options: cut prices, or make your platform so sticky that leaving is more painful than staying.
Nvidia chose both. The margin reset is the price cut. Hugging Face is the stickiness.
Let's anchor the earnings before we argue about what they mean, because the numbers are genuinely staggering.
| Metric | Value | Context |
|---|---|---|
| Q2 FY2027 revenue | $96.2 billion | Up 106% year-over-year⁴ |
| FY2028 revenue guidance | +70% growth | Analysts had modeled ~45%⁵ |
| ACIE customer segment | $40.3 billion | Up 138% YoY — non-hyperscaler demand⁶ |
| Gross margin (now) | ~75% | Heading to 72–73% next year⁷ |
| Post-earnings stock move | +7.2% premarket | Snapped a 4-quarter losing streak⁶ |
| Hugging Face price | $12.9 billion | ~86x its ~$150M revenue¹ |
Two numbers in that table deserve more ink than the rest.
First, the 70% forward guide. For the prior four quarters, Nvidia beat estimates and the stock still dropped the next day. This time it held — because for the first time, management gave Wall Street a longer runway. CFO Colette Kress said revenue will grow 70% in fiscal 2028. Huang, never one for modesty, said demand is "much greater than 70%" and that the only ceiling is supply.⁶ That's a man telling you the constraint isn't demand, it's physics and TSMC's factory floor.
Second, the ACIE number — $40.3 billion, up 138%. This is the quiet, under-reported story inside the story. ACIE stands for "AI Clouds, industrial, and enterprise" — the customers outside the hyperscaler club. It's Nvidia's answer to the single most damaging bear thesis: that its entire business rests on five companies that are all building their own chips. Huang said it himself: "This time last year, one lab alone was driving the build-out. Today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel."⁶ The customer base is genuinely widening. That's real, and it's bullish.
But don't let the width of the base distract you from the shape of the moat.
The consensus take this morning is pure relief: "Nvidia proved the AI boom isn't slowing." Every headline says some version of that. And sure — the numbers support it.
But the consensus is missing the second-order signal, which is where the actual money gets made or lost.
The gross-margin reset is the single most important data point Nvidia has released this year — and almost nobody is talking about it.
Huang called it "ripping the Band-Aid off."⁷ Here's the translation: Nvidia has "absorbed cost increases" (read: TSMC, memory, and HBM are all charging more), it has "repriced products in the marketplace" (read: it gave customers a break), and as a result margins will slide from 75% to 72–73% next year.⁷
Now, a 75% gross margin falling to 72–73% still makes Nvidia one of the most profitable hardware companies in human history. Apple would kill for those numbers. But direction matters more than level. A company choosing to give up margin is a company that sees a competitive threat before the market does. If you genuinely believed your customers had nowhere else to go, you'd raise prices, not lower them.
This is what the bulls don't want to admit: the $12.9 billion Hugging Face deal and the 2–3 point margin give-back are two halves of the same decision. Nvidia is defending against the moment its biggest customers stop being customers. Hugging Face gives it the open-source developer ecosystem — millions of engineers who build on Nvidia's hardware because the models are free and the tooling is familiar. That ecosystem is the counterweight to OpenAI and Google and Amazon going in-house.
Think of it this way: if every frontier lab builds a custom chip, Nvidia's moat shrinks to whoever is left. Hugging Face is Nvidia buying the rest of the market — the long tail of startups, enterprises, and independent developers who will never build their own silicon and who currently treat Nvidia GPUs as the default. That's not a growth engine. That's an insurance policy with an 86x premium.
And by the way — the deal also hands Nvidia a back door into cloud computing, which it quietly abandoned when it scaled back DGX Cloud a year ago.¹ Hugging Face already helps developers run models on rented compute. If Nvidia's customers sign $18 billion in cloud commitments and then don't use all of it (the company has pledged to cover tens of billions in cloud deals), owning Hugging Face gives Nvidia somewhere to offload the unused capacity.¹ It's the rare acquisition that hedges three risks at once: chip defection, cloud re-entry, and stranded compute.
There's a third headline this morning that's getting buried under the Nvidia euphoria, and it might matter more than either of the first two.
Washington is reportedly weighing a fresh round of tariffs on semiconductors — and this time the scope is broader. Politico reported Thursday, citing eight people familiar with the discussions, that the Trump administration is considering duties on an expanded range of tech products made alongside chips: laptops, data-center servers, and gaming hardware.⁸ The White House's only comment was a tell: "Reshoring semiconductor manufacturing is a top priority."⁸
Nvidia just told you it's absorbing higher input costs and resetting margins down. Now layer a tariff on servers and laptops on top of that. The company that's giving back 2–3 points of margin to hold market share does not have a lot of room left to eat a tariff. Something gives — either consumer prices go up, or margins compress further, or the reshoring timeline becomes a capex problem.
This is the macro risk the relief rally isn't discounting. Nvidia's quarter was clean. The political weather around it is not.
Here's where we stop narrating and start advising. Three concrete moves, no "consider diversifying" hand-waving.
1. Re-price your own "moat" assumptions — including the one in your portfolio. If you own Nvidia (or the semis more broadly), ask yourself one question: am I holding this because of the 70% growth, or because I believe the moat is durable? Those are different theses. The 70% is priced in. The moat is now being explicitly defended with $12.9 billion and a margin give-back — which means it was under more pressure than the headlines admit. If you can't answer why the moat holds for five years, trim to the position size that lets you sleep.
2. Watch the margin, not the revenue, next quarter. Revenue is going to beat. It always beats. The number that matters now is gross margin. If it prints below 72%, the market will finally understand what Huang just told it in plain English. If it holds at 73%+ despite tariffs and input costs, the moat is real and you stay long. Set your alert for "NVDA gross margin" and ignore the revenue headlines.
3. Position for the tariff catalyst, not against it. A semiconductor/server/laptop tariff is a reshoring trade. That's a tailwind for domestic fabrication, onshore packaging, and anyone selling "made in America" compute — and a headwind for the consumer-electronics margin stack Nvidia just told you is already thin. The contrarian play isn't shorting Nvidia. It's rotating a slice of your tech exposure toward the companies that benefit from Washington forcing chips back onshore. Do it before the Politico report becomes a signed order.
Every bull case has a shadow. Here's the honest version.
The Hugging Face deal could still die — or get blocked. It's a reported agreement, not a signed one. Business Insider explicitly said talks "could still atomize."¹ And a chip behemoth buying the largest open-source model repository is a competition-regulator's dream target, especially in a political climate where "open vs. closed" AI is already a live Washington fight. If it falls through, the moat-defense thesis loses its flagship.
Open-source might not save the moat — it might commoditize it faster. The uncomfortable irony: Hugging Face thrives on open-source models that run on anyone's silicon. If Chinese models like Moonshot's Kimi K3 keep matching U.S. labs at a fraction of the cost, open-source becomes the thing that breaks Nvidia's pricing power, not protects it. Nvidia is buying the platform precisely because open source is the one force it can't control by selling better chips.
The 70% guide is a promise against a supply chain Nvidia doesn't own. TSMC faces constraints; memory chips are in short supply.⁶ Huang admitted the ceiling is supply, not demand. If TSMC's next-gen capacity slips, or HBM allocation gets re-prioritized by a geopolitical shock, 70% becomes aspirational overnight — and the market that just rewarded the visibility will punish the miss twice as hard.
A 100% semiconductor tariff isn't priced anywhere. Trump has already floated "approximately 100%" tariffs on chips.⁸ The January 25% tariff on certain AI chips was a warm-up. If the broadened round lands at anything close to the floated number, the consumer-electronics and server businesses Nvidia just said it's defending get structurally repriced — and no amount of open-source moat fixes a demand shock.
Nvidia didn't report a bad quarter — it reported a brilliant one. But the two moves around it — a $12.9 billion open-source acquisition at 86x sales, and a self-inflicted gross-margin cut — tell you what management won't say out loud: the era of Nvidia as an unchallenged growth story is quietly giving way to Nvidia as a defended franchise. That's a better company and a trickier stock. Own the moat if you believe it; don't own the momentum.
The king isn't dead. He's just started building walls — and you don't build walls when the horizon is empty.
All claims verified against Silver-tier sources (CNBC, TechCrunch) and reputable financial/tech outlets (Fox Business, Invezz). Each source URL was scraped and confirmed accessible. Reuters and Forbes were paywalled and discarded. The Hugging Face acquisition is reported (The Information) and not yet company-confirmed — labeled as such throughout. Last verified: August 27, 2026.
You don't build a wall when the horizon is empty — and you don't cut your own margin when the monopoly is safe. 🎯
