A gold rush produces two kinds of winners. A handful of prospectors strike it rich, and one merchant sells every prospector the shovels, picks, and denim, no matter whose claim pays out.
The prospectors get the movie deals. The merchant gets the steady money.
For three years, the artificial intelligence (AI) boom has followed that script closely. OpenAI, Anthropic, and Elon Musk’s AI outfit trade blows over whose chatbot writes better code, and every new release arrives with a chart claiming the crown.
You have probably seen the scoreboard version of this story. One week a lab tops a leaderboard, the next week a rival posts a bigger number, and investors scramble to figure out which camp to back.
Behind nearly all of those models sits the same supplier. Its chips train them, its chips run them, and more and more, its money helps pay for them.
On Monday, Sept. 21, that supplier said the quiet part out loud. “Grok 4.7 has landed,” Nvidia’s official X account posted, adding that the company was “proud to support the team with NVIDIA accelerated computing.”
The post congratulated SpaceXAI on “its most capable model yet for coding and knowledge work.” It reads like routine vendor cheerleading, until you look at who else Nvidia is cheering for.
Why investors should pay attention to Nvidia’s Grok 4.7 post
Grok 4.7 is the newest model from SpaceXAI, the AI arm of SpaceX (SPCX) after the rocket company absorbed Musk’s xAI.
The model costs $2 per million input tokens and $6 per million output tokens and runs “twice as fast, at half the price of comparable models,” according to SpaceXAI.
More Artificial Intelligence:
- Google just disclosed something troubling about its AI
- White House drafts bold cybersecurity plan for tech’s next wave
- Bank of America hikes S&P 500 target after AI trade cracks
Musk had promised the model would “surpass all existing models,” reported The Standard. The scorecard is mixed. Grok 4.7 beat OpenAI’s GPT-5.6 Sol on the CursorBench 4.0 coding test, 46.3% to 41.7%, but trailed it on DeepSWE v1.1, 71.0% to 72.7%, Unite.AI noted.
That split result is the point. Nobody has locked up the lead, and Nvidia has arranged its business so it doesn’t need anyone to.
Nvidia shares closed at $227.16 on Sept. 21, up 2.2% from the prior session, according to Investing.com.
Nvidia owns a stake in all 3 top AI startups
Most suppliers pick a favorite customer. Nvidia bought a seat at every table in the frontier-model fight instead.
Here is where the chipmaker has money on the line across the AI race:
- OpenAI: A $30 billion stake, reported Yahoo Finance
- Anthropic: Up to $10 billion, also according to Yahoo Finance
- SpaceX, which now owns xAI: 122.8 million shares worth about $21 billion as of June 30, according to Nvidia’s 13F filing, as TheStreet reported on Nvidia’s SpaceX stake
- Safe Superintelligence: $5 billion, Yahoo Finance confirmed
I added up the three stakes in the labs fighting over the coding crown and got approximately $61 billion. In my analysis, that is about what Nvidia now earns in a single quarter, since the company earned $59.7 billion in its fiscal second quarter.
The overlap goes deeper than equity. SpaceX agreed in May to rent all of its Colossus 1 data center, more than 220,000 Nvidia graphics processing units (GPUs), to rival Anthropic, reported Data Center Dynamics.
“No one set off my evil detector,” Musk said of the deal, according to the same report. So Nvidia’s chips inside Musk’s building now power one of Musk’s competitors, and Nvidia holds stock in both.
Why Nvidia wins the AI race no matter who leads
Every lab that wants to stay in the race must keep buying compute, and nearly all of it runs on Nvidia hardware. That shows up in the numbers.
Nvidia’s second-quarter revenue hit $96.2 billion, up 106% from a year earlier, with $89 billion coming from its data center unit, according to Nvidia. The company guided to $108 billion for the current quarter.
Related: Zuckerberg, Musk, and Huang take key stand on huge AI issue
“Now, compute is revenue,” CEO Jensen Huang said in the same release.
I ran that guidance against the calendar. At $108 billion over a roughly 91-day quarter, Nvidia is booking about $1.2 billion a day.
Compare that with SpaceX’s biggest new AI hosting contract, worth $1.11 billion a month starting Dec. 1, reported Yahoo Finance. The shovel seller takes in about as much in one day as one of its best customers earns from a major deal in a month.
Wall Street mostly sees the lab stakes as affordable. Nvidia’s roughly $70 billion in AI investments is “easily manageable given its ability to generate as much as about $470 billion of free cash flow” over 2026 and 2027, Bank of America analyst Vivek Arya said, according to Yahoo Finance.
What Nvidia’s everywhere strategy means for your 401(k)
If you own an S&P 500 index fund or a target-date fund, you almost certainly own Nvidia already. Nvidia’s market value is about $5.5 trillion, which makes it one of the heaviest weights in any broad U.S. stock fund.
That means you don’t need to guess whether Grok, ChatGPT, or Claude wins the coding race. Your retirement account is already tied to the company selling to all of them.
The catch is that one bet on many labs is still one bet on a single trend. Nvidia’s direct sales remain “highly concentrated,” with one customer accounting for as much as 16% of revenue, according to its latest quarterly filing with the Securities and Exchange Commission (SEC).
Some of that demand is also shorter-term than the headlines suggest. SpaceX described its compute arrangements “as roughly 90-day commitments followed by a 90-day exit period,” CFO Bret Johnsen said, according to Yahoo Finance.
Put simply, a compute deal can unwind in about six months. If the labs trim orders, Nvidia would feel it long before any model loses a benchmark.
Nvidia’s practice of supplying chips to companies it also owns has drawn scrutiny, too. Chipmakers, hyperscalers, and AI firms are “linked through financing arrangements that can be opaque and difficult to value, leaving the financial system exposed if expectations of future profits fall short,” General Manager Pablo Hernández de Cos said in a Sept. 10 speech, according to the Bank for International Settlements.
The Nvidia risk to watch after Grok 4.7
Nvidia’s next test arrives with its third-quarter report in November. The number that matters most will be whether the frontier labs keep lifting their orders, since each one is also a customer Nvidia helped fund.
Musk wants to take the fight to orbit as well, and has said he is highly confident SpaceX will launch Nvidia AI computers into space next year, as TheStreet reported on Musk’s SpaceX and Nvidia plans.
For a merchant who sells to every prospector, a rival striking gold is good news. The day to worry is the day the prospectors stop digging.
Until then, Nvidia will keep posting congratulations to whoever ships next.
Related: Huang just doubled his chip forecast and blocked a regulator
