Nvidia is using two quarters of gross margin to buy certainty for the next two years.

After the market closed on August 26, Nvidia's stock price initially fell by more than 2%, but then rebounded halfway through the conference call, closing up more than 4%. Between the drop and the rise, the market was digesting the same figure.

This figure is for next year. For the first time, Nvidia has given revenue guidance a full year in advance, projecting 70% growth for fiscal year 2028, while the market consensus was only around 45%. Jensen Huang added the second half: without supply constraints, this figure would be much higher. The ceiling for demand is invisible; the ceiling for revenue is determined by production capacity.

The costs are clear. To boost supply, Nvidia openly paid up, lowering its gross margin guidance to 71% and increasing its procurement commitments from $119 billion to $279 billion in one quarter to lock in memory. Sacrificing gross margin and costs in exchange for increased capacity and revenue.

Demand has reached a crossroads; it's 100% likely to collide with 70%.

The reason for this deal now is that demand has reached a crossroads. Customers are forecasting nearly double Nvidia's growth ; however, due to supply constraints, Nvidia can only plan for 70% of its needs. With 100% demand and 70% supply, the difference is what Nvidia is vying for.

So for the first time, it gave guidance a full year in advance, projecting revenue growth of 70% for fiscal year 2028, far exceeding the market's previous estimate of 45%. Huang added the second half: "Without supply constraints, this figure would be much higher."

The short-term outlook is also very strong. In the third quarter, Nvidia expects revenue of $108 billion, fluctuating by 2%, reaching $100 billion for the first time in its history, representing a year-on-year increase of nearly 90% and a quarter-on-quarter increase of $11.8 billion, almost all of which comes from data centers.

The evidence of demand is everywhere. The cloud industry has over $2 trillion in orders on hand, and the top five hyperscale vendors have nearly $800 billion in capital expenditures this year, with plans to reach $1.3 trillion next year. Amazon alone has pledged to deploy another 2 million GPUs.

Where the demand comes from is also worth taking a look. Of the $89 billion in data center revenue this quarter, hyperscale customers contributed $48.7 billion, a year-on-year increase of 102%; AI cloud, industrial and enterprise customers contributed $40.3 billion, with a growth rate of 138%, growing even faster. Nvidia is growing a second wave of demand beyond the giants.

The price of this transaction is written in 74%.

The price of this transaction is stated in the profit margin guidance. The median gross margin for the third quarter was 74%, one percentage point lower than the 75% for the current quarter and also lower than the market expectation of 75%. In the fourth quarter, it is expected to reach 71% to 72%.

Kress stated bluntly that the memory market is experiencing extreme pricing conditions, with price increases exceeding expectations. Huang Renxun put it even more bluntly: wafers, HBM storage, and data center power are all in short supply; there is no overcapacity in the industry.

Two other forces are also pushing down gross margins. Vera Rubin is ramping up production, and the yield and cost of the new platform in the early stages are dragging it down in the short term; operating expenses are also rising, and non-GAAP expenses are expected to reach about $9 billion in the third quarter.

The decline in gross margin is due to the money Nvidia paid for this deal. Nvidia could have easily raised prices and passed the memory cost on to customers, preventing the margin from dropping. But instead, it chose to absorb the loss itself.

The 279 billion yuan procurement commitment serves as proof of this transaction.

The proof of payment is that $279 billion purchase commitment. It jumped from $119 billion in the previous quarter to $279 billion, a surge of $160 billion in just one quarter.

Management put it bluntly: the increase will primarily come from memory. These commitments lock in HBM and DRAM capacity for fiscal years 2028 and 2029. Nvidia has essentially pre-ordered memory capacity from the spot market.

This is more assertive than any statement expressing optimism about the memory market. Storage manufacturers talk about price increases, but that's just talk; Nvidia has secured its memory supply for the next two years with a $279 billion contract. Its multi-year partnership with SK Hynix paved the way for this.

Some on the sell-side have already interpreted this as an industry signal. Guojin Securities commented that AI is ushering in a new major cycle for storage. Nvidia's move to lock in production capacity ahead of schedule is tantamount to stamping this cycle on its head.

The hallmark of a monopolist is the willingness to proactively reduce profits.

Only Nvidia in the entire industry dared to make this deal. The mark of a monopolist is not how high the price can be, but the ability to decide when and by how much to concede.

Nvidia's financial strategy this quarter prioritizes volume over price. A two-point drop in gross margin is a trade-off for securing production capacity over the next two years. Their focus is on delivering the 70% growth that can be achieved; the current two-point margin difference is secondary.

More importantly, it split the profit-sharing into two parts. One part goes to the upstream, locking in the storage manufacturers' production capacity with high prices so they have the money to expand production; the other part goes to the downstream, absorbing the memory cost itself so that cloud providers can still make a profit in the computing power business.

This is the true meaning of "supply-constrained growth." The gross margin that the market is concerned about is just one cost item in this deal; what Nvidia is eyeing is whether the entire AI industry chain can work together to make money.

This deal is a gamble on the next-generation platform taking over on schedule. Vera Rubin, which just went into mass production this month, will contribute about 20% of data center revenue in the third quarter. Jensen Huang says its revenue potential per gigawatt is $40 billion, a significant increase over the previous generation. This is the confidence behind Nvidia's willingness to sacrifice profits.

The cost of this transaction is hidden in a distorted balance sheet.

This deal didn't come without a price. It pushed Nvidia significantly from an asset-light chip company towards an asset-heavy one.

The most telling sign is in the cash flow statement. Second-quarter free cash flow was $21.3 billion, more than halved from the first quarter's $48.5 billion. Accounts receivable increased by $22.3 billion, inventory increased by $5.8 billion, and prepayments increased by $5.5 billion, indicating that expansion requires Nvidia to advance increasingly more funds.

Even more significant are the ecosystem investments. At the end of the second quarter, Nvidia held approximately $93.9 billion in market-listed securities and unlisted investments, a substantial increase from the beginning of the year. It has also partnered with six institutions to leverage $500 billion in third-party capital to build AI infrastructure.

The credit market understood this cost before the stock market. The day before the earnings report, Morgan Stanley issued its first credit rating for Nvidia, giving it only a "neutral" rating, citing the increasing use of its balance sheet to finance its AI ecosystem; five-year CDS saw its largest single-day increase in history before the earnings report. As one of Societe Generale's strategy directors bluntly put it: "When looking at companies like this, you have to look at the CDS as well."

Nvidia hasn't been idle when it comes to shareholder returns either. This quarter, it repurchased and paid out $26 billion in dividends, and still has $99 billion in repurchase authorizations on hand. It's doing all three things simultaneously: locking in production capacity, investing in its ecosystem, and rewarding shareholders.

The market is conflicted about this trade, with prices falling initially and then rising.

The after-hours chart showing a decline followed by a rise reflects the market's indecisiveness regarding this trade. First look at the gross profit margin: sell; then look at the growth: buy.

Morgan Stanley points out a sobering fact: for the past four quarters, Nvidia's stock price has fallen the day after its earnings report, and "better-than-expected" has long been the default scenario. This time, the turnaround is due to genuine growth.

The disagreement on the sellers' side centers on price. Citigroup offers $300, Morgan Stanley $288, and Goldman Sachs $285, all significantly lower than the market average of $317. While 96% of institutions maintain a buy or overweight rating, no one dares to make a definitive statement regarding a potential memory price increase.

In this deal, Nvidia used two quarters' worth of gross margin to buy what it valued most: the certainty of growth. With demand doubling and supply only able to meet 70%, it chose to lock in supply first and then discuss profit margins.

The key to success lies in the fourth quarter. If the gross margin can stabilize around 71% and the recovery to 72% to 73% materializes, this deal will be a win. Nvidia has traded two quarters of profit margin for certainty over the next two years.

The conditions for falsification are equally clear. Memory price increases spiraled out of control, and attempts to correct them failed; downstream demand couldn't support the 70% commitment, turning the 279 billion yuan procurement reserve from ammunition into an inventory burden; further ahead, OpenAI's self-developed chips, Google's TPU, and Huawei's Ascend are gradually eroding its pricing power on the inference side. Whether the gross profit margin can stabilize around 71% in the fourth quarter is the first observation window.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.