𝐀𝐫𝐞 𝐰𝐞 𝐰𝐢𝐭𝐧𝐞𝐬𝐬𝐢𝐧𝐠 𝐭𝐡𝐞 𝐞𝐧𝐝 𝐨𝐟 𝐭𝐡𝐞 “𝐣𝐮𝐬𝐭 𝐫𝐞𝐧𝐭 𝐆𝐏𝐔𝐬” 𝐞𝐫𝐚?

NeoClouds emerged as the heroes of the global chip shortage.

NeoCloud offered what the giants couldn’t, i.e.,

⬩Speed.
⬩Flexibility.
⬩Lower costs (up to 85% less).

But a new McKinsey & Company analysis suggests the easy wins are over & the “Bare Metal” model is becoming a trap.

𝐓𝐡𝐞 𝐑𝐞𝐚𝐥𝐢𝐭𝐲 𝐂𝐡𝐞𝐜𝐤

⬩Margins are razor-thin
⤷ Real profits after costs can be as low as 14-16%.

⬩Depreciation is brutal
⤷ The value of a GPU fleet can drop by 50% in just five years.

⬩The Paradox.
⤷ To survive, they must move “up the stack” into software and services.
⤷ But doing so puts them in a direct war with their biggest customers: The Hyperscalers.

♦️The path forward isn’t volume.
✅ Its value.

⬩Sovereign Compute ➜ Managing sensitive, local data.
⬩Specialisation ➜ Niche workloads like ultra-low latency inference.
⬩Start-up Symbiosis ➜ Embedding deeply with the next generation of AI giants.

➜ Hardware gets you in the game.
➜ It’s the Software that keeps you there.

Differentiation is the new scale.

What’s your take?

Will NeoClouds evolve into specialised powerhouses or fade like the Cloud 1.0 startups?