Welcome to AI Collision đ„,
In todayâs collision between AI and our world:
- Several weeks of AI company deep dives
- Microsoft inks a big deal with a small AI factory
- Microsoft first, who comes next?
If thatâs enough to get the billion-dollar deals inking, read onâŠ
AI Collision đ„
Artificial intelligence doesnât float in the cloud it lives, physically, in factories.
Not smoke-stack factories, but often windowless, warehouse-sized supercomputers requiring more electricity than entire cities.
These are AI factories, and theyâre fast becoming the backbone of the twenty-first-century economy.
Right now, the world is building compute infrastructure at a speed and scale unseen since the creation of the modern internet.
Goldman Sachs estimates AI infrastructure spending will top US $1 trillion by 2030, while McKinsey forecasts that global data-centre power demand will triple to more than 1,000 terawatt-hours annually (roughly what Japan consumes each year).
A single gigawatt-scale site can cost US $50â60 billion to construct, and hyperscalers have already committed more than US $250 billion in 2025 alone toward expansion.
Every chatbot query, every image generator, every digital brain being trained consumes physical resources… energy, memory, bandwidth, and land.
The average large-language-model (LLM) training run now uses ten million times more compute than it did in 2012, and itâs still accelerating. As models scale, they need gigawatts of power, kilometres of fibre, and oceans of cooling water just to stay alive.
Thatâs what I’m going to look at in more detail with you for the remainder of this year.
I’m calling it the AI Factory Fortunes series, and I’m going to look at the lesser-known companies building the scaffolding of intelligence.
Every week, weâll spotlight at least one lesser known, key player in the AI build out. From data-centre builders and chip-host operators to AI power providers and infrastructure innovators.
These will all be the companies transforming AI from abstract code into physical infrastructure.
The AI build out isnât a story about ChatGPT and Gemini and the other big AI models from Google, Meta, OpenAI. Itâs about supply chains, substations, hardware, and physical infrastructure.
That tangible pieces of a puzzle that’s building our world for the next 1,000 years.
Itâs the construction of the worldâs digital foundation… the industrialisation of intelligence.
And somewhere inside that trillion-dollar (and eventually quadrillion dollar) buildout are the companies (and opportunities) that could define the next generation of wealth creation.
To kick things off we’re looking at an up and coming AI cloud company, Nebius (NASDAQ:NBIS).
Nebius â The Cloud Upstart Fueling Microsoftâs AI Ambitions
In this first installment, we look at Nebius (NASDAQ: NBIS), an under-the-radar cloud provider that just landed a whale of a contract with Microsoft.
Nebius Group, a relatively new AI infrastructure company headquartered in Amsterdam, made headlines after inking a multi-year agreement to deliver dedicated GPU capacity to Microsoft worth $17.4Â billion through 2031
That figure could rise to $19.4Â billion if Microsoft adds more capacity over the contractâs term.
To put it in perspective, this single deal is larger than all projected Nebius revenues through 2027.
No wonder Nebiusâs stock soared nearly 50% on the news. The deal will see Nebius providing Microsoft access to over 100,000 of Nvidiaâs latest AI chips housed in Nebiusâs new Vineland, New Jersey data center.
In essence, Microsoft is renting Nebiusâs GPU-packed facilities to bolster its own AI capacity â a creative workaround to satisfy sky-high internal AI demand while keeping Microsoftâs Azure cloud capacity available for its paying customers.
Why would Microsoft turn to Nebius?
The answer lies in the unprecedented scramble for AI compute.
With ChatGPT-style services and Copilot features appearing across the board in Microsoft products, Microsoft needs heaps of GPU horsepower.
But its data centers arenât built overnight, and Nvidia chips are in short supply.
Enter Nebius… a âAI cloudâ specialist that offers plug-and-play access to massive GPU clusters.
By tapping Nebius (and other so-called âneocloudâ upstarts), Microsoft can secure tens of thousands of Nvidia GPUs quickly, effectively outsourcing part of its own massive AI factory buildout to meet surging demand.
According to reports, Microsoftâs Nebius deal alone secures roughly 100,000 top-tier Nvidia GB300 chips for projects like building LLMs and AI assistants.
In other words, Nebius is an external assembly line for Microsoftâs AI efforts.
Nebius pitches itself as an âAI-nativeâ cloud platform built for heavy-duty AI workloads. It designs much of its own hardware and software, aiming to give AI developers turnkey compute and storage tailored for machine learning.
Landing Microsoft as a client is a massive validation for Nebiusâs model. Not only does the contract bring in a dependable revenue stream (Nebius essentially has $3.5Â billion per year guaranteed from 2026 onward), but it also boosts Nebiusâs clout in the market.
Essentially, Microsoftâs money helps Nebius build more AI âfactory floorâ square footage, which Nebius can then leverage for other customers too.
The CEO hinted this is just the first of several long-term contracts with leading AI labs and big tech, suggesting Nebius could snag additional hyperscale partners.
In the broader âAI factoryâ supply chain, Nebius represents the new breed of specialised AI suppliers ensuring the big names have enough raw compute to feed the AI boom.
Much like a parts manufacturer in a gold rush, Nebius benefits from selling picks and shovels (in this case, GPU capacity) to the miners (AI developers).
Of course, it’s still a risky play, and the ability to now execute on these deals remains. Nebius must actually build and deliver all that capacity on time.
But if it succeeds, Nebius will be deeply embedded in Microsoftâs AI pipeline for years, and possibly in other hyperscalersâ plans soon.
Nebius shows how even smaller-cap players can have a huge impact in the AI era.
Becoming an essential cog in Microsoftâs AI machine, Nebius has transformed virtually overnight from obscure cloud provider to a crucial âAI factoryâ contractor.
When the AI gold rush hits full stride, it may be the pick-and-shovel suppliers like Nebius that see their fortunes forged
The Media Is Wrong Again
They said Northern Rock was safe.
They said inflation was âtransitory.â
Now theyâre saying Britain should brace for recession.
But history shows⊠when the headlines scream âpanicâ, the real opportunity is often hiding in plain sight.
One contrarian investor believes the UK may be first to profit from this shift
Capital at risk.
Boomers & Busters đ°
AI and AI-related stocks moving and shaking up the markets this week. (All performance data below over the rolling week).
Boom đ
- D-Wave (NASDAQ:QBTS) up 15%
- Allegro Microsystems (NASDAQ:ALGM) up 5%
- Micron (NASDAQ:MU) up 5%
Bust đ
- Palantir (NASDAQ:PLTR) down 4%
- IBM (NYSE:IBM) down 4%
- Nvidia (NASDAQ:NVDA) down 5%
From the hive mind đ§
- A massive data centre deal coming from Down Under with Nvidia getting in on the action to launch several big data centres worth around $73 billion across Australia.
- It started with Sam Altman saying ChatGPT would allow for “erotica” it ended with them putting an AI mental health and wellbeing council in place.
- You know what solves the AI energy problem? I do. Nuclear fusion energy. And with AI we may be closer to that solution than ever before.
Artificial Polltelligence đłïž
Not a poll today, but a question…
With our look at the AI Factory Fortunes companies, we’ll be diving into some of the lesser-known names of the buildout of the AI boom.
Is there a company/stock that you’re hoping we’ll look at? Something small, under the radar that you know about that you’d love for us to look at? What’s your favorite AI stock right now that you think is a market sleeper?
Let us know via the comments below… and who knows… maybe you’ll see it appear in our deep dives over the next several weeks!
Weirdest AI of the day
ChatGPTâs random quote of the day
“The purpose of software engineering is to control complexity, not to create it.”
â Pamela Zave
Thanks for reading, and donât forget to leave comments and questions below,
Sam Volkering
Editor-in-Chief
AI Collision
