Two AI chip startups on opposite sides of the world announced major funding rounds on August 3, 2026, underscoring how aggressively investors are backing alternatives to Nvidia's dominance in artificial intelligence hardware. London-based OLIX closed a $312 million Series B at a $3.3 billion valuation, while South Korea's DeepX saw its value surge to $2.2 billion in a new funding round, according to Bloomberg.

For more context on this story, see our ongoing AI hardware coverage.

OLIX: A Token Factory Built From Scratch

OLIX, a two-year-old London startup founded by James Dacombe, raised $312 million at a $3.3 billion valuation in a round reported by Electronics Weekly. The financing attracted a striking mix of strategic and financial backers: chip designer Arm, quantitative trading firm Hudson River Trading, and investment fund Fundomo participated, alongside angel investor Reed Hastings, the co-founder of Netflix. Existing investors — including Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court, and Transition — all increased their commitments.

What sets OLIX apart is its insistence on designing every layer of its systems. Rather than buying off-the-shelf processors, the company designs the chips, the lasers, and the optical network that connects them, delivering its hardware as complete racks built for the most demanding AI inference workloads.

OLIX's pitch rests on a manufacturing metaphor. "A datacenter is a factory whose product is the token," the company argues, noting that producing a single token requires hundreds of operations, each placing different demands on hardware. In a conventional factory, every stage gets a machine built specifically for it. In today's AI datacenters, by contrast, every stage runs on the same general-purpose chip — an approach OLIX believes has reached its efficiency limits.

DeepX: Korea's Edge-AI Challenger Gears Up for IPO

On the same day, Bloomberg reported that South Korean AI chip designer DeepX reached a $2.2 billion valuation in its latest funding round. DeepX, which focuses on edge-AI processors — chips that run artificial intelligence directly on devices rather than in distant data centers — has been steadily building toward a public listing.

The company has already hired Morgan Stanley to arrange funding ahead of a planned initial public offering, Bloomberg previously reported, and Reuters confirmed in April 2026 that DeepX was preparing a public share offering. DeepX has also partnered with Hyundai to develop robots powered by generative AI, pointing to the automotive and robotics sectors as key demand drivers for its silicon.

DeepX's surging valuation reflects broader investor enthusiasm for Korea's homegrown AI chip industry, which the government and major conglomerates have backed as a matter of strategic sovereignty.

A Wider Wave of AI Silicon Investment

The OLIX and DeepX rounds are not isolated events. They fit a pattern of capital flooding toward companies that believe they can carve out territory Nvidia does not yet fully own — particularly in inference, the phase of AI where trained models respond to user queries, which is rapidly becoming the largest cost center for AI deployment.

In March 2026, fellow Korean startup Rebellions raised $400 million in funding explicitly aimed at leading a domestic revolt against Nvidia chips, as reported by SiliconANGLE. Across the Atlantic, Etched AI landed a $300 million Series C at a $10.3 billion valuation for its specialized inference chips, and AMD continued pushing its Instinct GPU line into the datacenter. The common thread is a conviction that the next phase of the AI boom will reward hardware optimized for running models — not just training them.

The optical-interconnect angle that OLIX is pursuing is particularly buzzy. Startup Eliyan raised $145 million in a Series C earlier this year for technology that uses light rather than electricity to shuttle data between chips, a bottleneck that grows more severe as models scale. OLIX's decision to integrate lasers and networking into its own racks places it squarely in that camp.

Why Inference Is the New Battleground

Training the largest models requires enormous clusters of high-end GPUs, a market Nvidia still commands overwhelmingly. But inference is where the volume — and the recurring cost — lies. Every ChatGPT query, every Claude conversation, every Gemini search runs through inference, and the economics of those operations determine whether an AI product is profitable.

Companies like OLIX argue that running every inference step on a general-purpose GPU is wasteful when each step has different computational needs. Building purpose-built silicon for distinct phases of token generation could dramatically lower the cost per query, which is why investors are willing to place billion-dollar bets on startups that have yet to ship at scale.

The risk, of course, is that Nvidia is not standing still. The company continues to iterate its own architectures and software ecosystem, and the moat created by its CUDA platform remains formidable. Startups betting on custom inference hardware must prove not only that their silicon is faster or cheaper, but that they can build the software stack and customer relationships to make it usable.

What the Numbers Signal

The simultaneous announcements from London and Seoul send a clear message about the state of AI hardware investment: capital is global, the sums are enormous, and the strategic logic has shifted from "who can train the biggest model" to "who can run models most efficiently."

With OLIX valued at $3.3 billion after just two years and DeepX approaching a public listing at $2.2 billion, the market is pricing in a future where Nvidia does not hold a monopoly on the silicon that powers artificial intelligence. Whether that future arrives on schedule remains the trillion-dollar question.

Sources

  • Electronics Weekly, "London AI chip startup OLIX raises $312m" (August 3, 2026)
  • Bloomberg, "South Korea AI Chip Designer DeepX's Value Surges to $2.2 Billion in Funding" (August 3, 2026)
  • Bloomberg, "AI Chip Firm DeepX Hires Morgan Stanley for Funding Before IPO"
  • Reuters, "Korean AI chip startup DeepX prepares public share offering" (April 2026)
  • SiliconANGLE, "Rebellions lands $400M in funding to lead the South Korean revolt against Nvidia chips" (March 2026)

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →