Pull to refresh
Logo
Cloud firms race to acquire AI inference optimization startups

Cloud firms race to acquire AI inference optimization startups

Money Moves

Nebius closes its $643 million Eigen AI deal, the latest in a wave of inference-talent buyouts.

June 16th, 2026: Nebius closes the Eigen AI deal

Overview

Updated Jun 16

Training an AI model is a one-time cost. Running it for millions of users, every day, is the bill that never stops. On June 16, 2026, Amsterdam-based cloud company Nebius closed a $643 million deal for Eigen AI, a 20-person U.S. startup whose only product is squeezing more answers out of each chip.

That price works out to about $32 million per employee. It is one of several deals in which a few cloud providers are buying up the small teams that know how to cut the cost of running AI. Whoever controls that cost controls the margins of the whole industry.

Why it matters

The cost of running AI models is now the industry's bottleneck, and a handful of cloud firms are buying up the talent that controls it.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

$643M
Eigen AI deal value
Cash-and-stock price Nebius paid to acquire the inference startup.
$32M/employee
Implied price per Eigen AI staffer
Eigen AI had about 20 employees at the time of the deal.
$399M
Nebius Q1 2026 revenue
Up from $50.9 million a year earlier as AI cloud demand surged.
684%
Nebius Q1 revenue growth
Year-over-year jump reported for the first quarter of 2026.

Voices

Curated perspectives — historical figures and your fellow readers.

Ever wondered what historical figures would say about today's headlines?

Sign up to generate historical perspectives on this story.

People Involved

Organizations Involved

Timeline

August 2024 June 2026

6 events Latest: June 16th, 2026 · 3 months ago
Tap a bar to jump to that date
  1. Nebius closes the Eigen AI deal

    Latest Acquisition

    Nebius completes the $643 million acquisition, folding Eigen's technology into Token Factory and opening a Bay Area hub.

  2. Nebius reports surging revenue

    Financial

    First-quarter 2026 revenue hits $399 million, up 684% year over year, with positive adjusted EBITDA.

  3. Nebius agrees to buy Eigen AI

    Acquisition

    Nebius announces a $643 million cash-and-stock deal for Eigen AI. Its shares rise about 12%.

  4. Inference clouds raise big rounds

    Funding

    Fireworks AI raises $250 million at a $4 billion valuation; Together AI had raised $305 million earlier in the year.

  5. Nvidia buys up inference startups

    Acquisition

    Nvidia absorbs optimization teams including OctoAI, Deci and CentML, treating inference software as a strategic layer.

  6. Nebius spun out of Yandex

    Corporate

    Yandex's non-Russian assets are restructured into Nebius, a Nasdaq-listed AI cloud company led by Arkady Volozh.

Scenarios

1

Nebius hits its 2026 revenue target

Likely Resolves by Mar 15, 2027

Discussed by: Nebius management guidance; coverage in Futurum and The Motley Fool

Nebius has guided to $3 billion to $3.4 billion in 2026 revenue, with Token Factory and the Eigen AI technology as growth drivers. If demand for managed inference holds and the Meta and Nvidia agreements ramp, the company lands inside or above that range. A miss would suggest the inference bet is taking longer to pay off than promised.

2

Another cloud provider buys an inference startup

Likely Resolves by Jun 16, 2027

Discussed by: The Next Web and tooldirectory.ai, tracking the consolidation wave

Nvidia, CoreWeave and Nebius have all bought optimization or inference teams. With inference cost now the main lever on AI margins, rivals such as CoreWeave, Together AI or Fireworks AI have a clear incentive to do the same. The trigger would be a publicly announced acquisition of an inference- or model-optimization startup by a major AI cloud provider.

3

Eigen AI's founder leaves Nebius

Unlikely Resolves by End of 2027

Discussed by: Standard acquisition-retention risk noted in M&A coverage

Acquired founders often depart once earn-out and integration periods pass, especially when a 20-person team is absorbed into a large company. Ryan Hanrui Wang is set to run Nebius's new Bay Area hub. If the integration sours or he starts a new venture, his exit would signal that Nebius bought technology more than a lasting team. Continued tenure would suggest the opposite.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

1993-2000

Cisco's acquisition-led growth (1993-2000)

During the early internet boom, Cisco bought dozens of small networking startups instead of building every product itself. It used acquisitions to assemble a full stack of routers, switches and software as demand exploded.

Then

Cisco filled product gaps quickly and became the dominant supplier of internet plumbing.

Now

Buy-to-build became a standard playbook for infrastructure firms in fast-growing markets, though some deals were later written down.

Why this matters now

Nebius is using the same strategy in AI infrastructure: acquire specialized capability to move faster than rivals can develop it.

2016-2019

Intel's AI chip acquisitions (2016-2019)

Intel bought AI chip startups Nervana in 2016 and Habana Labs in 2019, each for amounts reported in the hundreds of millions to roughly $2 billion, to catch up in machine-learning hardware.

Then

Intel gained teams and designs aimed at competing with Nvidia in AI compute.

Now

Results were mixed; Intel later shut down the Nervana line, showing that buying talent does not guarantee winning a new layer.

Why this matters now

It is the cautionary case for Nebius: acquiring a specialized team is only worth $643 million if the technology and people stay and deliver.

2024-2025

Nvidia's inference startup buying spree (2024-2025)

Nvidia acquired a run of optimization startups, including Lepton AI, OctoAI, Deci and CentML. The CentML deal was reported at up to $400 million. Each team specialized in making AI models run faster and cheaper on Nvidia chips.

Then

Nvidia pulled scarce optimization talent in-house and tied it to its hardware.

Now

The deals helped establish inference software as a distinct, acquirable layer of the AI stack.

Why this matters now

Nebius's Eigen AI purchase follows the same logic: buy the small team that controls chip efficiency rather than build it slowly.

Sources

(6)