← All companies

Chips / Company profile

Positron

What Positron does

Positron is a U.S.-fabricated AI inference hardware company focused on making transformer inference more cost- and energy-efficient than GPU-based approaches, and on enabling large-context, memory-bound inference workloads. The company ships “Atlas” as a transformer inference server/inference accelerator system today, and is building next-generation custom silicon and systems—“Asimov” (custom AI accelerator chips) and “Titan” (a next-generation inference system designed for very large memory and context)—as the roadmap to replace more of the traditional GPU stack with proprietary components over time. Positron positions its differentiation around memory-first architecture and performance-per-watt/per-dollar improvements, targeting data centers and inference operators that are constrained by power availability, heat/power densities, and memory/capacity limits rather than only raw compute throughput. The company also emphasizes compatibility and deployability via an inference software stack and the use of standard model ecosystems, with Atlas exposed via Positron’s inference engine and supported by a developer portal for OpenAI-compatible inference endpoints.

News

Company record

Feb 04, 2026 · Official · Positron press release PDF (via Web Summit Qatar materials)Positron announces $230M Series B at $1B+ valuation to scale energy-efficient AI inference; co-led by ARENA Private Wealth, Jump Trading, and Unless

Positron announced an oversubscribed $230M Series B at a post-money valuation exceeding $1B, naming co-leads and strategic/participating investors and reiterating the roadmap from Atlas today to Asimov/Titan. The release also includes claims about memory capacity/bandwidth targets and a tape-out/production timeline for Asimov.

Feb 04, 2026 · Reporting · TechCrunchPositron raises $230M Series B to take on Nvidia (exclusive coverage)

TechCrunch reported on Positron’s $230M Series B and described investor participation including QIA, and framed the funding as a step toward competing in AI inference chips and systems.

Date not disclosed · Reporting · EE TimesPositron COO/CTO Thomas Sohmers discusses shipping Atlas systems and customer traction; FPGA-based first/second-generation Atlas; CEO appointment context

EE Times reported that Positron has been shipping FPGA-based LLM inference systems and referenced deliveries to customers and a multi-million-dollar order narrative, alongside discussion of its Atlas/roadmap approach.

Date not disclosed · Reporting · Tom’s HardwarePositron introduces performance/per-watt claims for Atlas vs Nvidia H200 (power envelope and tokens/sec framing)

Tom’s Hardware reported Positron’s claims comparing Atlas inference performance and power consumption against Nvidia’s H200, including a headline-style statement about power usage and tokens/sec for a referenced workload.

Date not disclosed · X · Mitesh Agrawal (Positron)Mitesh Agrawal announces $51.6M Series A led by Valor, Atreides, and DFJ Growth; names additional investors and partner/angel descriptors

On LinkedIn, Positron’s CEO posted that the company raised a $51.6M Series A led by Valor Equity Partners, Atreides Management, LP, and DFJ Growth, with additional investment from Flume Ventures, Resilience Reserve, 1517 Fund, and Unless, along with named individuals described as part of those investor groups.

Source map · 2 recurring channels · 24 references

Still resolving: Newsroom · Official X

Funding

Latest disclosed valuation$1B
Tracked capital$305.1M3 sourced rounds

What it builds

Atlas

Atlas is Positron’s transformer inference server/inference system that the company says is shipping today. Positron’s product page lists system-level claims including performance-per-watt and performance-per-dollar comparisons versus Nvidia systems, and provides hardware/software specifications (including Positron Inference Engine).

source ↗
Titan

Titan is Positron’s next-generation inference system that the company describes as coming in 2027. Positron positions Titan as a memory-first platform for very large context windows and multi-trillion-parameter workloads, powered by four Asimov chips and designed for air-cooled operation and scale-out.

source ↗
Asimov

Asimov is Positron’s custom AI accelerator silicon. The company describes Asimov as a memory-first chip designed around transformer inference’s memory bandwidth/capacity bottlenecks, with specifications such as per-chip memory capacity, realizable memory bandwidth, TDP, data types, and host/interface details.

source ↗

Key metrics

Milestones & partnerships