Nvidia on the Edge: Two Customers and Alibaba’s Chip Could Rattle it's AI Revenues?

Nvidia’s latest SEC filing reveals that two unnamed buyers — “Customer A” and “Customer B” — together accounted for 39% of Q2 revenue (23% and 16%, respectively), up sharply from the same quarter a year earlier. At the same time, Chinese cloud giant Alibaba has developed and is testing a new AI inference chip designed to handle a wide range of inference workloads — a move that could, over time, reduce Chinese buyers’ reliance on Nvidia hardware. Both items are real, current developments investors should monitor because each raises plausible near- and medium-term downside scenarios for Nvidia revenue and margins.

1) Risk — rising customer concentration

Nvidia’s recent 10-Q / quarterly filing discloses that Customer A represented ~23% of Q2 revenue and Customer B ~16%, meaning nearly 40% of the quarter’s revenue came from just two buyers (likely fellow MAG 7 companies). That’s materially higher than the comparable quarter a year earlier, when those two customers accounted for about 14% and 11% respectively. The company’s public results still look strong (Nvidia reported ~$46.7B in Q2 revenue), but the line in the filing highlights a concentration risk that is often buried in footnotes.

Why this matters to investors

• Revenue volatility: If one large buyer pauses purchases, delays deployments, or shifts to a competitor (or to internally developed solutions), Nvidia’s top-line could swing meaningfully quarter-to-quarter because a small number of customers now represent a very large share of sales.

  
  

• Pricing & bargaining power: Large customers can extract better pricing, longer payment terms, or exclusive arrangements when they represent a large share of a supplier’s revenue. That can compress gross margins if negotiations change.

  
  

• Geopolitical / policy vulnerability: If these big buyers are based in markets subject to export controls or political pressure (China is often discussed in the media in this context), regulatory shifts could disproportionately affect sales. Analysts commonly point to cloud providers as likely candidates for big-ticket buys — historically the cloud giants have been large buyers of accelerators — which adds a policy dimension to the concentration risk.

  
  

Practical scenarios to consider

• Short term (next 1–2 quarters): A single major customer pauses orders for inventory reasons → revenue down materially vs. consensus in a quarter; management calls this out on the next earnings call.

• Medium term (6–24 months): One or both large customers diversify suppliers (or build custom infra) → slower revenue growth for Nvidia’s Data Center segment and possible margin pressure.

• Downside tail: Loss of one top buyer combined with greater discounting to retain business could hit EPS more than top-line headlines imply.

2) Risk — Alibaba’s new AI inference chip

Multiple outlets report that Alibaba has developed a new, domestically produced AI inference chip aimed at supporting a broader set of inference workloads for its cloud customers and is currently testing it — part of a wider push in China to reduce dependence on foreign (Nvidia) chips under export constraints. Reports emphasize the chip is focused on inference, not on the high-end training accelerators that power the largest foundation-model training jobs.

Why this matters to Nvidia investors

• Addressable segments differ, but overlap exists. Alibaba’s chip is described as an inference-focused accelerator. Nvidia’s business includes both training (high-end GPUs) and inference (a broader set of accelerators including H20/H100 families and lower tier GPUs). If Chinese cloud providers begin routing more inference workloads to local chips, Nvidia could lose a portion of inference revenue even while maintaining a lead in top-end training hardware.

  
  

• Domestic manufacturing & policy tailwinds. Because the new chip is reportedly manufactured domestically (not by TSMC) and aligns with Chinese industrial policy to build supply-chain independence, adoption may be accelerated within the Chinese market irrespective of raw performance parity. That makes substitution risk more credible inside China than in other regions.

  
  

• Ecosystem & software lock-in remain big hurdles. Nvidia’s CUDA ecosystem, software stack, and broad third-party optimizations remain a major competitive advantage. Even if Alibaba’s silicon is competitive on cost or certain throughput measures, the burden of porting models and validating production systems can slow migration — so substitution is plausible but unlikely to be instantaneous.

  
  

Putting the two risks together — how they reinforce each other

Customer concentration becomes a bigger strategic vulnerability when those concentrated customers can realistically switch to alternative suppliers. If the large unidentified buyers driving ~39% of revenue are (or include) major Chinese cloud consumers, and if Alibaba’s inference silicon gains traction inside China as reports indicate, the combination of high concentration + localized alternatives magnifies downside risk in that market segment. Even if only a portion of inference demand shifts, the magnitude of revenue tied to a couple of buyers makes the earnings impact asymmetric and worth monitoring closely.

Investor action checklist, what to watch next

1. Watch Nvidia’s customer disclosures & comments on earnings calls — management may provide more color about who the big buyers are, contract length, and how sustainable current purchase levels are. (Start with the next 10-Q and the earnings call transcript.)

   
   

2. Track Alibaba chip adoption signals: product announcements, availability in Alibaba Cloud regions, benchmark publications, and partner integrations (model frameworks, SDK support). Early commercial rollouts are the clearest sign of meaningful revenue impact.

   
   

3. Monitor pricing / ASP trends for Nvidia’s Data Center segment. Sharp ASP deterioration could indicate discounting to retain large buyers or a shift in mix toward lower-margin channels.

   
   

4. Listen to cloud providers’ comments (Microsoft, Amazon, Meta) during their reports for signs of inventory builds or supplier diversification — these firms are widely believed to be large buyers of AI accelerators. Analysts and filings often hedge-name them as potential “customers A/B.”

   
   

5. Consider portfolio sizing & hedges. If you own Nvidia, decide how much weight you’re comfortable with given a plausible downside from customer attrition or regional substitution. For new positions, factor concentration and geopolitical substitution risk into target prices and time horizons.

   
   

Bottom line

Nvidia’s business remains enormously profitable and the secular AI demand story is intact, but two concrete, near-term developments — a sharp increase in revenue concentration among two unnamed buyers and the reported emergence of a domestically produced Alibaba inference chip — materially raise the company’s exposure to idiosyncratic and region-specific risks. For investors this is not an immediate sell signal, but it is a legitimate reason to tighten monitoring, reconsider position sizing, and watch upcoming filings and product adoption metrics closely.

For latest Nvidia analyst ratings click here.