Why $300B AI deals still happen in a world of NEO Clouds

Another week, another headline. Oracle signs a $300B AI infrastructure deal. Microsoft expands its GPU footprint. Google and AWS double down on AI superclusters.

The hyperscaler arms race is accelerating – and it’s all in pursuit of ‘AI dominance’.

But here’s the paradox. At the same time, a new breed of cloud providers – the so-called NEO clouds – is rising fast. They’re leaner, cheaper, and built for agility. So why are enterprises still pouring hundreds of billions into hyperscaler AI deals?

Let’s unpack it.

What is a NEO cloud?

NEO clouds aren’t just smaller versions of AWS or Azure. They’re fundamentally different. Think CoreWeave, Lambda Labs, Scaleway, Vultr – platforms designed for speed, simplicity, and specialisation.

They’re built around GPU-heavy AI workloads, edge compute and sustainable data centres, transparent pricing, developer-friendly APIs, open-source tools and minimal vendor lock-in.

In short, they’re the antidote to hyperscaler complexity. And they’re booming – especially among startups and mid-sized firms who are priced out of traditional cloud models.

So, why the mega-deals?

If NEO clouds are faster, cheaper, and more flexible, why are hyperscaler AI deals still happening at such massive scale? A few reasons.

Scale & capital intensity

Training frontier AI models – think trillion-parameter systems – requires tens of thousands of GPUs, specialised networking, and billions in upfront investment. Only hyperscalers have the capital and infrastructure to deliver that kind of scale.

NEO clouds are brilliant at provisioning GPUs quickly. But, they’re not yet equipped to build and operate global AI superclusters.

Enterprise trust & integration

For large enterprises, AI isn’t a standalone project. It’s embedded into ERP systems, databases, analytics platforms – all of which already run on hyperscalers. Migrating mission-critical workloads to a niche provider introduces risk. Hyperscalers offer continuity and integration.

Global reach & compliance

Enterprises need global delivery, multi-region access, compliance with local laws, and data sovereignty guarantees. Hyperscalers have the footprint for it, while NEO clouds tend to be regional or workload-specific.

Ecosystem & partnerships

Deals like Oracle’s aren’t just about compute. They’re about strategic lock-in – securing exclusive partnerships with AI labs, embedding AI into core software products, and shaping ecosystems that will define the next decade. That kind of influence is beyond the reach of smaller players.

Where NEO clouds fit in

None of this makes NEO clouds irrelevant. In fact, they’re essential disruptors.

They offer affordability for startups, flexibility for developers looking to avoid vendor lock-in, and even innovation in niches like sustainable cloud or edge AI.

They thrive where hyperscalers are slow, expensive, or overly complex. In many ways, they’re the pressure valve preventing a total monopoly on AI compute.

The reality will be a multi-cloud AI future

The rise of NEO clouds doesn’t cancel out hyperscaler mega-deals; the two are actually complementary.

Hyperscalers will continue to power enterprise-scale AI with global infrastructure, compliance, and ecosystems.

NEO clouds will fuel innovation, speed, and affordability for those who need agility.

I believe the winning strategy will be to go multi-cloud by design. Yes, use hyperscalers for scale and trust – but at the same time, leverage NEO clouds for flexibility and disruption. The balance between the two will define how organisations compete in the AI-driven decade ahead.

NEO clouds are rising fast. But hyperscaler mega-deals aren’t slowing down, because at the bleeding edge of AI – capital, compliance, and global scale still rule the game.

The future isn’t either/or. It’s both.