Nvidia's Twin Engines: $500,000 H200s Head to China While the Balance Sheet Bankrolls AI's Buildout

Nvidia delivers H200s to China, backs OpenAI data center with $105B guarantee, and expands AI financing platforms ahead of Q2 earnings.

The news cycle around Nvidia has rarely been this crowded. On Wednesday, the first batch of H200 processors touched down in China, with ByteDance and Tencent each receiving roughly 10,000 units, according to the Financial Times. That delivery marks the end of a months-long import blockade and the beginning of a far larger pipeline: as many as ten Chinese companies have been cleared to bring in up to 100,000 H200s apiece, with Nvidia preparing approximately 500,000 units in total for the mainland market.

There is, however, a catch that tempers the headline number. Most of the licensed chips must remain in Hong Kong rather than on the Chinese mainland — and the territory currently lacks the data-center capacity to put them to work. The more powerful B200 remains entirely off-limits for export.

The China development lands in the middle of a pivotal stretch for the stock. Nvidia reports quarterly earnings on August 26, and Bank of America reaffirmed its buy rating and $350 price target the day before the delivery news broke. Analyst Vivek Arya calls Nvidia an "unmatched growth franchise," pointing to a price-to-earnings ratio of roughly 18 — a seven-year low for the shares.

That valuation gap reflects how far the business has run ahead of the stock. Revenue grew 85 percent to $81.6 billion in the first fiscal quarter of 2027, with the data-center segment surging 92 percent to $75.2 billion. Management has guided to around $91 billion for the current quarter, and Stifel's Ruben Roy expects the company to clear that bar when it reports.

A Financing Machine Takes Shape

While the China shipments grab headlines, a quieter transformation has been underway on the balance sheet. Nvidia has committed to a guarantee of up to $105 billion to back OpenAI's leasing of an SB Energy-developed data center in Ohio, and it is investing $1.5 billion directly in the SoftBank-owned energy arm. Reuters reports the guarantee was initially discussed at $250 billion before being trimmed to under $120 billion — evidence that the company is still calibrating how much risk it wants to carry.

The Ohio deal is hardly an outlier. In early August, Nvidia joined Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR in announcing plans to launch compute-financing platforms targeting more than $500 billion in third-party capital for AI infrastructure. Goldman has already approached insurers, asset managers and banks to anchor the investor base, per Reuters. Add in a commitment of up to $3 billion to Lancium, the energy infrastructure developer behind the Stargate campus in Texas — including a $2 billion investment for roughly a 20 percent stake — and a pattern emerges: Nvidia is not just selling chips, it is building the financial scaffolding that makes the demand for those chips possible in the first place.

The Market's Mixed Verdict

The stock's recent action suggests investors are still digesting what this dual role means. In Frankfurt, Nvidia shares traded at €188.02 on Wednesday, down 0.9 percent on the day. The weekly decline stands at 3.4 percent, though the stock remains about 4 percent above its 50-day average of €180.54 — a sign the near-term trend has not broken despite the pullback. The picture on the other side of the Atlantic is similar: Tuesday's close of €189.72 represented a 2.4 percent drop, with a 2.5 percent weekly loss and a 6.5 percent monthly gain. The shares sit 6.3 percent below their 52-week high of €202.50.

The broader semiconductor complex has been a drag. Samsung Electronics and SK hynix both suffered double-digit losses in their home markets, and the Philadelphia Semiconductor Index fell nearly 5 percent. Yields on ten-year US Treasuries have climbed to multi-decade highs, a headwind for growth names generally. Cathie Wood, for her part, says she avoids memory-chip stocks entirely, favoring alternatives like Cerebras that operate without high-bandwidth memory.

The Product Roadmap Continues

None of this has slowed the engineering pipeline. Nvidia is developing Nemotron 4, a new family of AI models designed to challenge leading open-source offerings globally — though Reuters notes final training is incomplete and no release date has been set. The company has also deepened its strategic partnership with SK Group around AI factories and next-generation memory technology.

The August 26 earnings report will be the moment of truth on two fronts: whether the growth trajectory holds despite geopolitical friction, and whether the massive financing commitments are already leaving marks on the income statement — or, as Nvidia frames it, are primarily a catalyst for future chip demand. The trimmed Ohio guarantee suggests the company itself is keeping a close eye on how much weight its balance sheet can carry.