Nvidia Wants to Be the Balance Sheet Behind AI — and Washington Is Starting to Ask Questions

Nvidia is buying Hugging Face for $12.93 billion, reaffirming ~70% fiscal 2027 revenue growth guidance as the DOJ probes its Groq chip deal.

Nvidia's ambitions no longer fit inside a chip fab. Within a matter of days, the company has moved to absorb one of the most widely used developer platforms in artificial intelligence, floated a revenue target that would have seemed fanciful a year ago, and found itself on the receiving end of a federal antitrust inquiry — all while its share price drifts sideways.

Start with the acquisition that says the most about where Nvidia thinks the value sits. The company is buying Hugging Face for $12.93 billion, one of the largest deals in its history. Roughly $11.9 billion goes to existing investors, with up to $1 billion set aside in stock-based retention packages for employees making the move. Crucially, Nvidia has pledged that Hugging Face will remain an open platform and that using Nvidia hardware will not become mandatory. Read one way, that is a shrewd public-relations play. Read another, it is an admission that a company of this size now competes on trust and neutrality as much as on raw performance. Whoever controls the industry's infrastructure cannot afford to look like a gatekeeper.

Growth Forecast, and a Shifting Bottleneck

Speaking at Goldman Sachs' Communacopia + Technology conference on Thursday, chief executive Jensen Huang reaffirmed guidance of roughly 70 percent revenue growth for fiscal 2027 and named cybersecurity as the next major AI application area, backed by partnerships with CrowdStrike, Cisco and Palantir. Nvidia is also running Palantir's Foundry and Nemotron technology through its own supply chain — testing its AI tools on itself before selling them into manufacturing, energy, healthcare and aviation.

Huang's more consequential claim, though, was that the bottleneck in AI buildout is migrating from silicon to financing. Nvidia, in other words, does not want its systems understood merely as depreciating hardware but as durable collateral — what Morgan Stanley has labeled "Balance Sheet as a Service." The bank lifted its price target to $300 while staying cautious on credit risk, projecting roughly $200 billion in lending exposure by the end of 2028, about $170 billion of it contingent liabilities tied to leases, residual-value guarantees and revenue-sharing arrangements.

Demand, by Nvidia's own account, shows no sign of cooling. The company put month-over-month growth in Grace, Blackwell and NVLink systems at 27 percent, and Huang described appetite as strong across cloud providers, enterprises, device makers and the so-called neoclouds. The geographic push is just as aggressive: 2 gigawatts of data-center capacity earmarked for Australia by 2027, attached to $80 billion in investment. In Southeast Asia, the provider Zankore secured a credit facility of up to $3.1 billion to build GPU and cloud infrastructure in Indonesia alongside Indosat Ooredoo Hutchison and Nokia, using Nvidia technology.

From the Data Center to the Desk

A second thread runs closer to home. In October, Lenovo and Acer will release the first Windows PCs built around the RTX Spark chip, which pairs a Blackwell graphics processor with a Grace CPU. Capabilities once confined to data centers are thus heading for laptops and desktops — the logical extension of a strategy in which dominating cloud infrastructure is not enough if the endpoint is left to rivals. Down the supply chain, Foxconn — described by Reuters as Nvidia's largest server manufacturer and a key Apple supplier — has guided for third-quarter performance ahead of market expectations, a sign of how far Nvidia demand now reaches.

Circular Financing and a Federal Question

The criticism Huang increasingly has to field concerns how all of this is paid for. The Bank for International Settlements warned in its annual report that circular AI financing represents a trillion-dollar systemic risk, and investors are asking whether Nvidia's stakes in CoreWeave, OpenAI and other customers amount to buying demand that is later presented as organic growth. Huang rejects the charge outright: the investments are "immaterial" relative to the business generated, he says, and the company only enters "sure-win" deals. That is persuasive as long as underlying demand is genuine — but the assurance alone does not shrink the balance-sheet exposure.

Meanwhile, the US Department of Justice is investigating Nvidia's chip-technology arrangement with Groq, valued at $17 billion, on suspicion the deal was structured to sidestep antitrust review. Nvidia did not acquire Groq; it secured non-exclusive technology rights and brought in key personnel including founder Jonathan Ross — a pattern critics call a "reverse talent acquisition." Any finding would more likely bring fines than an unwinding, yet the regulatory climate remains a genuine uncertainty.

A Stock That Won't Cheer

The market has been notably restrained about all of it. The shares trade at EUR 189.42, up 0.7 percent on the previous day, after shedding 4.5 percent across the prior seven sessions. That leaves the stock 6.5 percent below its 52-week high of EUR 202.50, set in May, though still up 18 percent year to date. On a shorter lens, the price sits just 2.4 percent above its 50-day moving average and 6.4 percent under that same high, with 30-day volatility at 40 percent and a twelve-month gain of 25 percent. Piper Sandler and Morgan Stanley have both set $300 targets, citing the Rubin architecture's promise of thirty times the throughput per megawatt versus Grace Blackwell Ultra, at falling token costs.

Politics adds another layer. Huang appeared at the G20 innovation ministers' meeting while the US lobbied other G20 members for looser copyright rules on AI training — not yet an Nvidia-specific decision, but a reminder of how tightly the company and regulators are now entangled. For investors, the question is no longer whether Nvidia grows. It is whether a pace of acquisitions and investments this relentless converts into substance rather than headlines — and whether supervisors and credit markets keep endorsing the construction.