Curious about today's AI digest?ai-tldr.dev

Daily Digest

Pomegra Startups

Olix Raises $312M Series B at $3.3B Valuation

**UK photonic chip startup Olix raised $312 million Series B at a $3.3 billion valuation** — closing Europe's largest semiconductor VC round, with ARM and Hudson River Trading joining, to build specialized inference chips using novel photonic interconnects.

Venture CapitalSemiconductorsMAJOR4 min read
Olix Raises $312M Series B at $3.3B Valuation

London photonic chip startup Olix closed a $312M Series B at a $3.3B valuation, with Arm and Hudson River Trading backing its inference chip bet on light-based interconnects.

Key Takeaways

  • Olix raised $312M at a $3.3B valuation on August 3, 2026, the largest semiconductor VC round raised by a European startup.
  • Its DX-1 chip uses light-based interconnects and ditches high-bandwidth memory to cut latency and supply chain exposure.
  • First customer delivery targets H2 2027; the company is hiring across five cities on two continents.

Lead

Olix, a London semiconductor startup founded in 2024, closed a $312 million Series B on August 3, valuing the company at $3.3 billion - more than triple its valuation from six months prior. The round drew in Arm Holdings, quantitative trading firm Hudson River Trading, infrastructure fund Fundomo, the UK government's Sovereign AI venture fund, and Reed Hastings, co-founder of Netflix. The company designs Optical Tensor Processing Units (OTPUs), chips that move data using light rather than copper, and is targeting the AI inference market with its first product, the DX-1, built specifically for large language models running at 100 billion parameters or more.

What Does Olix Actually Build?

The DX-1 is a decode accelerator targeting the output stage of large language models. That narrows its addressable market compared to general-purpose accelerators, but the company argues specificity is the point: by focusing on one workload, Olix optimizes its photonic interconnect architecture without the overhead of supporting training or smaller models.

The defining technical choice is how data moves between chips. Olix's "slow and wide" optical interconnect transfers information via light, cutting the latency and energy costs associated with copper chip-to-chip communication. The company also drops high-bandwidth memory from its design entirely - an unusual decision that reduces dependence on HBM suppliers and the advanced packaging processes that have created supply chain bottlenecks across the semiconductor industry. Olix claims DX-1 delivers more than 10,000 tokens per second per user with better throughput-per-watt than general-purpose chips. Those figures are the company's own. First customer access is scheduled for the second half of 2027.

Why Is Arm Investing in a Chip Startup?

Arm licenses the instruction set architectures that underpin most of the world's processors; Olix builds photonic inference silicon. The two are not direct competitors, but the relationship is still worth examining. Arm's stake signals it wants strategic exposure to photonic compute, not just licensing revenue from it. For Olix, the backing provides a credibility marker that its chips can integrate with Arm-based systems - important in a data center market where Arm architecture is growing its share.

Hudson River Trading's participation is less obvious. The quantitative trading firm has been building internal chip research capacity for years, and investing in inference accelerators fits a pattern of financial firms seeking hardware-level computational advantages. Whether HRT eventually becomes a customer is undisclosed.

The Valuation Question

At $3.3 billion, Olix is valued at more than ten times what most observers would assign a company with no chips in production and no disclosed revenue. The previous valuation, implied to have been well under $1 billion six months before the August close, suggests investors are pricing in either a sharp revision in expectations or a competitive dynamic that puts a premium on early photonic compute positions.

The surrounding context matters. Nvidia's grip on AI training hardware has pushed investors toward differentiated inference plays, and photonics - long viewed as too costly and complex for commodity compute - has attracted renewed capital as AI workloads grow. Valuations across AI chip startups expanded significantly from late 2024 onward, compressing the skepticism that would ordinarily accompany a two-year-old company with no customer shipments.

Olix is hiring across London, Bristol, Austin, Toronto, and San Francisco, covering silicon design, compiler development, photonics engineering, and systems integration. The appointment of Professor Nick McKeown, a Stanford computer networking authority, to the board adds credibility on the systems architecture side - but hardware delivery schedules are a separate matter from board composition.

Outlook

Olix has capital, high-profile strategic backers, and a technology thesis that distinguishes it from the crowded field of GPU alternatives. The questions that will determine whether the $3.3 billion valuation holds are straightforward: does DX-1 ship on schedule in 2027, and does measured performance match the claims? Every dollar in this round is a bet that it does. The next checkpoint is production silicon in customer hands. Until then, the $312 million buys time, not proof.

More Startup News