Kursol
AI Breaking News is an AI-generated alert, curated and reviewed by the Kursol team. When major AI developments happen, we break down what it means for your business.
Etched AI completed a $700 million Series D funding round at a $21 billion valuation on August 18, 2026, with quantitative trading firm Jane Street leading the investment and signing on as the company's first customer. The valuation doubled in just one month — Etched hit $10.3 billion in July — signalling a dramatic shift in how investors value non-NVIDIA AI chip companies. Jane Street is already running Etched's first rack of chips (a physical unit of server hardware) in production, making this the first real-world deployment of a direct NVIDIA rival for AI workloads.
For the first time in five years, NVIDIA faces a credible infrastructure alternative that customers are actively using — not testing, not piloting, but running live. That changes everything about how you negotiate compute costs.
Why Etched Matters When NVIDIA Looks Inevitable
For years, NVIDIA's dominance in AI chips felt like physics. The company sold the vast majority of the GPUs powering AI, set prices, and controlled supply. The Nvidia-funded OpenAI data centre deal announced yesterday cemented that position — NVIDIA financed the infrastructure and locked in exclusive supply. Public cloud providers built in NVIDIA dependencies because they had no alternative. Organisations followed suit.
Etched changes this calculus in one specific, powerful way: it is built for inference — the moment when your trained model answers real questions from real users. That's where most AI cost accumulates. Whilst NVIDIA's GPUs handle everything (training and inference), Etched chips are engineered specifically for inference, which makes them cheaper and faster at the thing most organisations actually do with AI.
Jane Street — a firm that makes money on precision and speed — deployed Etched in production. They would not run live trading on hardware that underperforms. That validation is worth more than any press release.
What This Means for Your Compute Budget
This is not yet a situation where you can walk away from NVIDIA entirely. But it is the first moment in which your operations team can legitimately evaluate alternatives for your inference workloads — the expensive ongoing cost, not the one-time training spend.
The maths changes immediately: if Etched can run your inference workloads for meaningfully less than NVIDIA-based services, that's material to your budget. Public cloud providers (AWS, Google Cloud, Azure) now have more room to negotiate NVIDIA prices because they can credibly say "we can offer Etched." NVIDIA's margins compress. Your costs stabilise or decline.
This is how vendor markets work when monopolies break: the first credible alternative changes pricing across the board, not just for customers who switch.
What to Do This Week
If you're evaluating a new AI infrastructure vendor: Ask whether they offer Etched as an inference option. If not, ask why — and ask when they will. The answer signals whether your cloud provider sees Etched as real competition or marketing noise.
If your team is mid-deployment: You don't need to migrate from NVIDIA today. But you should understand what fraction of your costs are inference vs. training. If inference is 60%+ of your spend (typical for live production systems), having an Etched alternative on your radar is strategically valuable. Talk to your cloud provider about roadmaps.
For finance teams: Model infrastructure costs with two scenarios: baseline (current NVIDIA-only pricing) and competitive (Etched enters your provider's offerings). The gap between these is your potential savings as the market matures.
This kind of vendor assessment is exactly what Kursol does for clients — understanding where real competition exists in AI infrastructure, where single-vendor lock-in remains, and where your team has actual negotiating leverage.
The Bottom Line
The NVIDIA monopoly in AI chips just cracked. It will take time for alternatives like Etched to reach price parity and availability across all cloud providers, but Jane Street running production workloads on Etched makes this no longer theoretical. Your infrastructure costs are about to become negotiable again.
If this development has you rethinking your AI infrastructure strategy, take our free AI readiness assessment to understand where you stand.
AI Breaking News is Kursol's rapid analysis of major artificial intelligence developments — focused on what actually matters for your business. Subscribe to our RSS feed to stay informed.