Is d-Matrix profitable in 2026?

d-Matrix does not disclose audited financials; it is private and VC-backed. Its Corsair chip entered full production on June 9, 2026, but no confirmed revenue or profitability figures have been publicly verified as of that date.

Who owns d-Matrix and how much funding has it raised?

d-Matrix is a private, VC-backed startup. As of November 12, 2025, it closed a $275 million Series C at a $2 billion valuation, bringing total funding to $450 million since its 2019 founding, per company and press releases.

Who is d-Matrix's biggest competitor?

Nvidia dominates with roughly 92% of AI-inference GPU market share as of February 2026 and paid about $20 billion in December 2025 to license Groq's chip tech and hire its team, directly entering d-Matrix's low-latency inference niche.

Could Nvidia crush d-Matrix?

Nvidia faces no structural block to entering d-Matrix's niche and already spent roughly $20 billion in December 2025 licensing rival Groq's technology and hiring its team, a direct move into the same inference market d-Matrix targets.

What does d-Matrix actually make?

d-Matrix builds AI inference accelerators using Digital In-Memory Compute (DIMC) architecture. Its Corsair platform, which entered full production June 9, 2026, is manufactured via TSMC's N6 node and Alchip, and the company claims 30,000 tokens/sec at 2ms/token for Llama3-70B in one rack.

How does d-Matrix compare to SambaNova and Cerebras?

SambaNova raised $1 billion at an $11 billion valuation in July 2026 and landed JPMorgan Chase as a customer; Cerebras completed an IPO in 2026. d-Matrix, valued at $2 billion as of November 2025, trails both in disclosed customer wins and valuation.

Semiconductor companies · United States

d-Matrix

d-matrix.ai ↗

Startup building in-memory-compute AI inference chips like Corsair and Raptor.

Losing its protection

Assessed 24 July 2026 · Evidence as of 24 July 2026

The structural protection is eroding.

On the recordSince 24 July 2026 · Checkable from mid 2026 · Every call on the record Checkpoint: First public hyperscaler production customer disclosure for Corsair

The call

d-Matrix has real chips shipping and unique memory-compute IP, but sits between a dominant Nvidia and richer rivals like SambaNova and Cerebras with no named hyperscaler customer yet.

d-Matrix designs specialized computer chips that process artificial intelligence work more efficiently by doing math inside memory rather than shuttling data back and forth. The company has actual silicon shipping in production starting mid-2026 with a customer called Corsair, which is harder to fake than vaporware, but Nvidia dominates AI chips with 92% of the market and just bought Groq's technology for $20 billion, giving the giant company an easy path into exactly what d-Matrix is building. The real question is whether d-Matrix can prove its technology works better in the real world and keep customers loyal while burning less cash than better-funded rivals like SambaNova and Cerebras, which have more money to spend on catching up.

Why

AI pressure: Nvidia's Groq-derived inference stack. Nvidia paid about $20B for Groq's core tech and could fold a rival inference stack into its dominant platform, squeezing d-Matrix's niche. Could AI do the core job? Only partly.

Bigger players: Nvidia. Nvidia already ships the GPUs running about 92% of AI inference workloads, and its CUDA software keeps customers locked in. Could a giant copy it? Yes, easily.

What protects it. Shipped, working silicon (Corsair in production) with proprietary DIMC and 3DIMC in-memory compute that Nvidia has not matched yet.

What would change this call

A named hyperscaler or frontier lab publicly discloses a Corsair production deployment within 12 months. Would show d-Matrix can convert its chip into real customer stickiness.
Nvidia launches a Groq-derived inference product targeting the same low-latency niche. Would show the incumbent closing d-Matrix's technical gap directly.
d-Matrix raises a Series D at a flat or down valuation versus the $2B November 2025 mark. Would signal weak commercial ramp and funding strain versus peers.

Key numbers

$275M raised, $2B post-money valuation (2025-11-12)
$450M total funding since 2019 (2025-11-12)
Nvidia holds about 92% of inference GPU share (2025-11-12)

Named risks

No publicly named hyperscaler or frontier-lab customer confirmed for Corsair as of June 2026. Severity: high
Nvidia has no structural barrier to entering d-Matrix's inference niche and already licensed Groq's tech for ~$20B. Severity: high
Better-funded rivals SambaNova ($11B valuation) and Cerebras (public) may outpace capital and customer acquisition. Severity: medium
Raptor accelerator and 3DIMC commercialization still unproven beyond validated test silicon as of March 2026. Severity: medium

Rivals, as read in July 2026

Rival	Direction	Evidence
Nvidia	contesting	Licensed Groq's tech and hired its team for ~$20B, Dec 24, 2025, entering d-Matrix's niche
SambaNova	losing to	SambaNova raised $1B at $11B valuation and landed JPMorgan Chase as customer, Jul 8, 2026
Cerebras	losing to	Cerebras completed IPO in 2026, ahead of d-Matrix's private $2B valuation status
Groq	gaining on	Groq's core IP/talent absorbed by Nvidia via $20B licensing deal, Dec 24, 2025, removing a direct peer

Questions people ask

Is d-Matrix profitable in 2026?: d-Matrix does not disclose audited financials; it is private and VC-backed. Its Corsair chip entered full production on June 9, 2026, but no confirmed revenue or profitability figures have been publicly verified as of that date.
Who owns d-Matrix and how much funding has it raised?: d-Matrix is a private, VC-backed startup. As of November 12, 2025, it closed a $275 million Series C at a $2 billion valuation, bringing total funding to $450 million since its 2019 founding, per company and press releases.
Who is d-Matrix's biggest competitor?: Nvidia dominates with roughly 92% of AI-inference GPU market share as of February 2026 and paid about $20 billion in December 2025 to license Groq's chip tech and hire its team, directly entering d-Matrix's low-latency inference niche.
Could Nvidia crush d-Matrix?: Nvidia faces no structural block to entering d-Matrix's niche and already spent roughly $20 billion in December 2025 licensing rival Groq's technology and hiring its team, a direct move into the same inference market d-Matrix targets.
What does d-Matrix actually make?: d-Matrix builds AI inference accelerators using Digital In-Memory Compute (DIMC) architecture. Its Corsair platform, which entered full production June 9, 2026, is manufactured via TSMC's N6 node and Alchip, and the company claims 30,000 tokens/sec at 2ms/token for Llama3-70B in one rack.
How does d-Matrix compare to SambaNova and Cerebras?: SambaNova raised $1 billion at an $11 billion valuation in July 2026 and landed JPMorgan Chase as a customer; Cerebras completed an IPO in 2026. d-Matrix, valued at $2 billion as of November 2025, trails both in disclosed customer wins and valuation.

Timeline

2026-07-08SambaNova raises $1B at $11B valuation, lands JPMorgan Chase as customer
2026-06-25Corsair wins 2026 AI Breakthrough Award (AI Processor Innovation Award)
2026-06-09Corsair enters full production, ships to hyperscalers, neoclouds, frontier labs
2026-03Pavehawk 3DIMC test chip validated within performance and power targets
2025-12-24Nvidia licenses Groq's chip tech and hires its team for about $20B
2025-11-17d-Matrix and Alchip announce 3DIMC 3D-stacked DRAM collaboration
2025-11-12d-Matrix closes $275M Series C at $2B valuation
2025-10-14SquadRack rack-scale architecture announced with Arista, Broadcom, Supermicro
2024-11-19Corsair unveiled, claims 30,000 tokens/sec at 2ms/token for Llama3-70B

Revision history

24 July 2026

First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.

In the same band

Wolfspeed Astera Labs Mythic