TensorWave

Founded in 2023 and based in Las Vegas, TensorWave is an AI cloud built exclusively on AMD Instinct GPUs, positioning itself as a deliberate alternative to Nvidia-dominated infrastructure. Led by co-founder and CEO Darrick Horton, the company deploys MI300X-class and later MI325X accelerators to serve AI training and inference workloads. In October 2024 it raised a $43M SAFE round—reported as the largest in Nevada startup history—led by Nexus Venture Partners, and in June 2026 closed a $350M Series B led by Magnetar and AMD Ventures at a $1.55B valuation to expand its global data-center footprint and deploy next-generation AMD Instinct GPU clusters.

It is one of the most prominent pure-play AMD GPU clouds, testing whether AMD's accelerators can be a credible at-scale alternative to Nvidia for AI compute.

tensorwave.com AI / GPU Cloud

United Statesprivate

At a glance

Total raised

$493M

last round series b

Headcount

51-200

approximate · public range

Headquarters

Las Vegas

United States

Founded

2023

3 years active

Status

Private

AI / GPU Cloud

Latest valuation

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Already subscribed? Sign in →

AI bull / bear

Bull case

TensorWave is the purest lever on the single most important question in AI infrastructure: whether AMD breaks Nvidia's monopoly. The June 2026 $350M Series B at $1.55B — co-led by Magnetar and AMD Ventures, taking total raised to $493M — roughly tripled the company's valuation in about a year and buys deployment of MI355X clusters aimed at exactly the workloads where AMD's silicon is strongest: memory-intensive training and high-throughput inference, where Instinct's HBM capacity advantage lets larger models run on fewer chips. The strategic geometry is unusually favorable. AMD needs credible at-scale cloud showcases and has capital, allocation, and engineering support to lavish on its flagship pure-play partner; customers need any alternative to Nvidia allocation queues and pricing; and TensorWave, founded only in 2023 and led by ex-Lockheed Skunk Works engineer Darrick Horton, has compounded through that gap — from a $43M SAFE in October 2024 (reported as Nevada's largest) to a $100M Series A to this round in under two years. ROCm's maturation removes the historical objection a quarter at a time. If AMD reaches even 15–20% of AI accelerator share, the neocloud that made itself synonymous with AMD compute inherits demand it never had to buy.

Bear case

TensorWave concentrates every neocloud risk and adds a single-vendor bet on top. Its entire fleet rides AMD's roadmap and ROCm's still-real software gap: every training-stack incompatibility, every kernel that ships CUDA-first, is friction TensorWave absorbs on behalf of a supplier it cannot influence. The cap table telegraphs the fragility — AMD Ventures co-leading is vendor financing with circular logic (AMD funds the company that buys AMD chips), and Magnetar is the same structured-capital playbook that levered CoreWeave, suggesting obligations senior to equity. Worse, TensorWave's differentiation is not exclusive: Oracle, Azure, and Vultr all deploy Instinct GPUs at scales TensorWave cannot match, so "AMD-first" is a marketing position, not a moat — if AMD succeeds, its chips are everywhere and TensorWave is just another mid-sized GPU landlord; if AMD stalls, TensorWave owns a depreciating fleet of second-choice silicon. At $493M raised against competitors deploying tens of billions, it is structurally subscale in a business where electricity, land, and cost of capital decide margins. GPU fleets depreciate in three to four years, neocloud consolidation is already the industry's own forecast, and a $1.55B valuation prices in AMD share gains that remain, for now, a thesis rather than a fact.

Generated by editorial-sourced · July 1, 2026Refreshes quarterly · see disclosures