A tiny Dutch startup called Euclyd has just landed Samsung as a backer in its bid to challenge Nvidia's grip on AI GPUs — but its inference chip won't ship until 2028. For UK businesses weighing whether to lock in expensive Nvidia capacity now or wait for cheaper competition, that three-year gap is the whole story.
What is the Concept
Euclyd is one of a growing wave of challengers targeting Nvidia's dominance in AI inference chips, the hardware that runs trained models rather than trains them. Samsung's backing signals real manufacturing credibility, not just a pitch deck. But a 2028 shipping date means this is a long-term supply story, not something that changes procurement decisions a UK business needs to make in 2026.
For businesses in London, Manchester, and Edinburgh currently budgeting in GBP for AI infrastructure, the practical question isn't "should we wait for Euclyd" — it's "how do we avoid overcommitting to today's pricing while a more competitive market is clearly forming.
Why It Matters Now (2025–2026 Context)
Nvidia's pricing power has driven up the cost of AI compute across UK enterprises throughout 2025, squeezing margins for companies running AI workloads at scale. News of credible challengers like Euclyd, alongside AMD and homegrown efforts, signals that this pricing power won't last indefinitely. Finance teams reviewing multi-year GPU or cloud AI commitments in 2026 now have a genuine reason to negotiate shorter contract terms rather than lock in long-term rates.
This matters most for mid-sized UK firms without hyperscaler-level purchasing power, who are the most exposed to Nvidia's current pricing and the most likely to benefit once real competition arrives.
How AI Is Changing This
Inference — running an already-trained model in production — is where most UK businesses actually spend their AI compute budget, far more than training. That's exactly the market Euclyd and similar challengers are targeting, because inference workloads are more predictable and easier to design specialised silicon for. As more inference-focused chips reach market from 2027 onward, expect a genuine price war that training-focused chips haven't triggered.
Here's the contrarian view worth stating plainly: most UK companies don't need to buy any GPU hardware outright at all. Renting AI compute through cloud providers, and staying contractually flexible, matters more right now than picking a hardware winner three years before it ships.
Real-World Examples
Consider a UK fintech running fraud-detection inference at scale. Locking into a three-year Nvidia-based contract in 2026 protects capacity but forfeits any pricing benefit if Euclyd, AMD, or another challenger disrupts the market by 2028. A UK retailer using AI for demand forecasting, by contrast, sized its compute needs modestly and uses flexible cloud-based GPU access — it can switch providers or chip generations as the market shifts, with far less lock-in risk. Samsung's own manufacturing scale is precisely why its backing of Euclyd is credible rather than speculative; it's the kind of partner that can actually get a chip into production.
The founder mistake here is assuming today's GPU market structure is permanent and building five-year infrastructure bets around it.
Practical Insights / Actions
UK businesses should run what we call a Compute Flexibility Check before any major AI infrastructure commitment: confirm contract length against realistic hardware refresh cycles, confirm whether workloads are inference-heavy (where new competition matters most) or training-heavy (where Nvidia's lead persists longer), and confirm there's an exit or renegotiation clause tied to market pricing shifts.
The hidden opportunity is negotiating leverage: UK procurement teams can now credibly cite emerging competition, including Samsung-backed entrants, when negotiating current Nvidia or cloud GPU pricing, even before any rival chip ships.
Future Outlook
Expect 2026 to be a year of positioning rather than disruption — Nvidia retains its lead, but credible challengers like Euclyd give UK buyers real negotiating leverage for the first time in years. By 2028, when Euclyd's chip is expected to ship, UK businesses that kept their AI infrastructure contracts flexible will be best placed to capture the pricing benefit of genuine competition.
Conclusion
The takeaway for UK businesses isn't to wait for Euclyd — it's to avoid rigid, long-term commitments while the AI chip market is visibly becoming more competitive. RP SoftTech helps UK SMEs design AI infrastructure and cloud strategies that stay flexible as the hardware market shifts, so today's spending decisions don't become tomorrow's overpriced liability.