Nvidia's Next Big Test: Can Ordinary Businesses Pick Up Where Amazon and Google Leave Off?
Nvidia earns most of its money from a handful of giant tech companies. Investors watching Wednesday's earnings want proof that thousands of smaller firms are finally starting to spend too.
Key points
- Nvidia reports fiscal second-quarter earnings on Wednesday, with analysts expecting revenue of $92.2 billion, nearly double a year earlier.
- Five or six hyperscalers account for roughly 55% of Nvidia's data centre sales.
- In Q1 2026, Nvidia's non-hyperscaler customers generated $37.5 billion in revenue, growing 31% quarter-on-quarter versus 12% growth from hyperscalers.
- Amazon, Alphabet and Meta all saw sharply lower or negative free cash flow in Q2 2026, raising questions about how much more they can spend.
- Nvidia recently signed letters of intent with six financial firms to unlock up to $500 billion in chip financing for smaller buyers.
Nvidia makes the specialised computer chips, called GPUs (graphics processing units), that power almost every major AI system running today. It has turned that position into a company worth $5 trillion. That wealth rests on a narrow base.
Why does customer concentration matter?
Lose one or two of Nvidia's biggest customers and the numbers move sharply.
The heaviest buyers are the hyperscalers: Amazon, Google and Microsoft, plus Meta and SpaceX, a group of five or six companies responsible for more than half of all data centre chip sales. Data centre chips now make up 94% of Nvidia's total revenue, so the maths are stark.
In Q1 2026, hyperscaler sales reached $37.9 billion. The rest of the market, which Nvidia labels ACIE (AI clouds plus industrial and enterprise customers), came in just behind at $37.5 billion. ACIE grew 31% from the previous quarter, comfortably ahead of the 12% posted by the hyperscaler group. That is the trend investors want to see accelerate.
Should investors be worried right now?
The cash flow picture at Nvidia's biggest customers gives them reason to pause.
Amazon and Alphabet both posted negative cash flow in Q2. Meta's cash generation fell more than 90% year-on-year. SpaceX and Tesla also reported negative free cash flow. Gene Munster, managing partner at Deepwater Asset Management, told CNBC: "The hyperscalers just can't give much more."
Nvidia shares fell 2.9% on Monday, the seventh consecutive day of losses and the longest losing streak since 2022, leaving the stock down 7.5% over that stretch.
What is Nvidia doing to broaden its customer base?
It is trying to make buying chips affordable for companies that cannot write a billion-dollar cheque.
Nvidia signed memoranda of understanding with six financial firms this month covering up to $500 billion in financing for businesses that want GPU access. The idea is to treat chips like real estate: assets that generate returns and can back a loan. We have been following this financing push since it first surfaced on 14 August, including a piece asking whether GPU resale values can actually support the pitch. Analyst Stacy Rasgon at Bernstein called the plan promising but thin on detail, telling CNBC it needs more specifics "to put some investors' minds at ease."
A new product line, the Vera Rubin GPU system, is also starting to ship. CEO Jensen Huang told developers at Nvidia's GTC conference earlier this year that he expects $1 trillion in sales from the current Blackwell and Vera Rubin generations through 2027. KeyBanc analysts, who rate the stock a buy, wrote over the weekend that ramping Rubin GPU shipments are the key factor that could push results above expectations.
| Metric | Q1 2026 actual | Q2 2026 estimate |
|---|---|---|
| Total revenue | ~$44B | $92.2B |
| Hyperscaler revenue | $37.9B | $43.6B (est. +83% YoY) |
| ACIE revenue | $37.5B | $43B (est. +149% YoY) |
| Data centre share | 92% | 94% |
What does this mean for smaller businesses and workers?
If the financing programme works, access to AI computing could get cheaper and more widely available. Right now, renting serious AI computing power is expensive because demand outstrips supply and only giant firms can buy in bulk. More lenders competing to fund chip purchases could, over time, lower the cost of building AI tools for mid-sized businesses.
The programme is still at the letter-of-intent stage. No binding contracts, no confirmed rates, no timeline.
The one honest takeaway: If you run a business and AI computing costs feel out of reach, watch how Nvidia's financing programme develops over the next two quarters. It is the most concrete attempt yet to bring GPU access within reach of companies that are not Amazon or Google.
