Four AI-chip startups raised Series B rounds in 2026 ranging from $135 million to $500 million β MatX, Positron, Fractile, and XCENA β all chasing a piece of the AI inference and training market Nvidia still dominates.
Semiconductor corporates led by Nvidia, AMD, and Samsung have participated in more than 60 startup financings of $100 million or more so far in 2026, in rounds collectively worth over $250 billion, according to Crunchbase News. Most of that total value traces to a handful of giant late-stage rounds. But a separate tier of Series B deals β smaller, earlier, and aimed squarely at Nvidia's inference and training hardware moat β tells its own story about where investors think the next generation of chip winners will come from. Four of the largest disclosed rounds in that tier, explained below.
2026 Semiconductor Startup Series B Announcements, Ranked by Size
MatX's $500 million round, reported in February, is more than double the size of any of the other three. All four rounds closed within a single six-month window, between February and May 2026.
| Attribute | MatX | Positron | Fractile | XCENA |
|---|---|---|---|---|
| Round size | $500M | $230M | $220M | $135M |
| Announced | Feb 24, 2026 | Feb 4, 2026 | May 14, 2026 | May 29, 2026 |
| Lead investor(s) | Jane Street, Situational Awareness | ARENA, Jump Trading, Unless | Accel, Founders Fund, Factorial | Atinum, IMM Investment |
| Post-money valuation | Not disclosed | $1B+ | Not disclosed | $570M |
| HQ | US | Reno, NV | UK | Santa Clara, CA / South Korea |
| Core bet | Custom chips for AI labs' models | High compute-per-watt inference | In-memory compute simulation-to-chip | Memory-centric computing (CXL) |
| Shipping status | Design finalizing, ships 2027 via TSMC | Atlas chip shipping now (Intel-fabbed) | Targeting customer delivery in 2027 | Mass production targeted late 2026 |
Sources: Bloomberg Law on MatX; StorageNewsletter on Positron; Yahoo Finance on Fractile; BusinessWire, via Rutland Herald, on XCENA. All figures as reported through each round's announcement date.
MatX: $500M to Build Chips for AI Labs' Own Models
MatX raised more than $500 million in a Series B reported by Bloomberg Law on February 24, 2026, led by Jane Street and Situational Awareness β the investment firm founded by former OpenAI researcher Leopold Aschenbrenner. Other named backers include chipmaker AMD Ventures, venture firms NFDG and Spark Capital, and Stripe co-founders Patrick and John Collison. MatX declined to disclose an exact valuation but said the company is now worth several billion dollars.
The capital is earmarked for securing scarce manufacturing capacity and memory components, and for pushing MatX One β the company's first chip β toward tapeout in under a year. MatX plans to finalize its design in 2026 and begin shipping through TSMC in 2027, targeting AI labs that want chips tuned to their own model architectures rather than general-purpose GPUs.
Positron: $230M at a $1B-Plus Valuation, With a Chip Already Shipping
Positron, based in Reno, Nevada, announced a $230 million Series B on February 4, 2026 at Web Summit Qatar, co-led by ARENA Private Wealth, Jump Trading, and Unless, with new participation from the Qatar Investment Authority, Arm Holdings, and Helena, per StorageNewsletter. The round valued the company at more than $1 billion, up from a $51.6 million Series A in July 2025.
Unlike the other three companies here, Positron already has a shipping product: Atlas, its first-generation inference chip, fabricated by Intel in the US. The company claims Atlas delivers roughly three times the compute per watt of Nvidia's H100 GPUs β a company claim, not an independently verified benchmark. Its next chip, Asimov, is on track to tape out in October 2026 for production in early 2027.
Fractile: $220M for In-Memory Inference Chips
UK-based Fractile, founded in 2022 by Oxford researcher Walter Goodwin, raised $220 million in a Series B announced May 14, 2026, led by Accel, Peter Thiel's Founders Fund, and Factorial Funds, with Conviction, Gigascale, O1A, Felicis, Buckley Ventures, and 8VC also participating, per Yahoo Finance. The company had raised $22.5 million earlier in 2026, and in February announced plans to invest Β£100 million in its UK operations over three years.
Fractile's approach is in-memory compute: chips that run calculations directly inside computer memory rather than shuttling data back and forth to a separate processor, which the company says can run inference 25 times faster and at 10% of the cost of current alternatives β again, an unverified company claim. Fractile is expanding its team across the UK, US, and Taiwan and targets data-center builders and AI companies for delivery starting in 2027.
XCENA: $135M at a $570M Valuation for Memory-Centric Computing
XCENA, a Santa Clara-based startup founded in 2022 by former Samsung Electronics and SK hynix engineers, closed $135 million (KRW 202 billion) in Series B financing on May 29, 2026, co-led by Atinum Investment and IMM Investment, bringing its total fundraising to $185 million at a $570 million valuation, according to the company's own BusinessWire release. Existing backers SBI Investment, Mirae Asset Capital, STIC Ventures, Wonik Investment Partners, SV Investment, and LB Investment returned, joined by new investor Korea Development Bank.
XCENA's product, MX1, is a computational memory controller that connects to the CPU via CXL and handles data-preprocessing and KV-cache management tasks directly inside the memory module during AI inference β a narrower, complementary bet than the other three companies' full custom chips. XCENA expects mass production of MX1 by the end of 2026, with revenue starting in 2027, and is expanding into Northern California to work more closely with hyperscalers.
What the headline misses
A wave of $100M-plus Series B rounds reads like proof that AI chip startups are displacing Nvidia. None of the four has a disclosed, named enterprise customer contract. Only Positron has a chip shipping to customers at all; the other three are still 2026β2027 stories on manufacturing partners, design tapeouts, and mass-production targets they have not yet hit. Every performance claim cited above β compute-per-watt, inference speed multiples, cost reductions β comes from the company itself, not an independent benchmark.
The money is real and the investor list is serious β sovereign wealth funds, Stripe's founders, Arm, and strategic chipmakers are all writing checks. But a Series B is still early-stage capital for a sector where fabrication, yield, and software ecosystem lock-in have sunk far better-funded chip challengers before. Nvidia's own corporate venture arm and rivals like AMD and Samsung backing more than 60 other AI and robotics startups this year, per Crunchbase News, is itself a sign that incumbents see more upside in hedging across many bets than in any one of these four being the breakout winner.
The Bottom Line
MatX's $500 million Series B is the largest of 2026's disclosed semiconductor startup Series B rounds, ahead of Positron's $230 million, Fractile's $220 million, and XCENA's $135 million β together $1.085 billion aimed squarely at Nvidia's grip on AI chips. Positron is the only one of the four with a chip already shipping; the rest are still racing toward 2027 production targets. Watch for Series C rounds to separate the companies that hit their manufacturing timelines from the ones still raising on design claims β Positron's own 5x markup into an $875 million Series C later in 2026 shows how fast that gap can open.
Track private AI and chip valuations across the market on the AI Valuations dashboard at Value Add VC.
Latest from the Pulse
Get VC data most people never see
β free to subscribe
Trace's notes on venture, AI and startups, a few times a week. Join 5,000+ subscribers. No spam.