Amazon, Microsoft, Google, Meta, and Oracle are tipped to burn through US$602 billion on infrastructure in 2026, with roughly 75% of that capex tied directly to the generative AI frenzy, funding servers, GPUs, optical networking, and the massive real estate footprints needed to rack it all.
Goldman Sachs reckons cumulative hyperscaler capex will hit $1.15 trillion between 2025 and 2027 - which is more than double the cash incinerated over the preceding three years.
But while the equities crowd sweats the widening AI monetisation gap, the tech heavies are simultaneously bankrolling the exact architecture engineered to strand those very assets.
A classical bit is strictly locked to being either a one or a zero; while a qubit bypasses this hardware gridlock via superposition, occupying both states simultaneously.
Combined with quantum entanglement, these rigs can chew through specific problem classes at speeds that make the heaviest GPU clusters look like dial-up.
Precisely why Microsoft's corporate vice president of Quantum, Zulfi Alam, going on the record with CNBC this week matters to the capital markets.
Alam bluntly stated that by 2029, machines inside commercial data centres will run calculations that no current system can handle.
Just twelve months ago, the Redmond giant refused to put a timeline on commercial viability.
This accelerated roadmap forces immediate, ugly questions about the depreciation schedules on the GPU-dense facilities being commissioned right now.
Hardware milestone
Microsoft's quantum play relies entirely on the Majorana 1 chip, which ranks as the first processor built on topological qubits.
This setup uses a topoconductor material that Microsoft insists can scale to millions of qubits on a single die.
Competing superconducting builds hit hard scaling ceilings that topological qubits are specifically engineered to bypass, handing Microsoft a structural edge over the blueprints chased by IBM and Google.
Over in Denmark, Microsoft is throwing cash at Atom Computing to build the Magne quantum system, representing a capital injection north of DKK 1 billion (A$223.5 million).
Magne slams together Microsoft's error-correction software with Atom's neutral-atom hardware, with switch-on expected by late 2026.
S&P Global analyst Ellie Brown pegs the timeline for commercial deployment between 2028 and 2032, with IBM, Google, and Amazon all riding parallel development tracks.
The same players on both sides of the trade
Microsoft torched through $34.9 billion on capex in a single quarter last year, a 75% year-on-year spike driven almost entirely by the scramble for AI compute.
Simultaneously, the software behemoth continues to run its largest quantum facility on earth.
All four major hyperscalers are now spinning up pay-as-you-go quantum cloud access, a setup built on the exact same commercial rails used to monetise AI data processing.
The underlying product swaps out, but the billing architecture doesn't blink.
AI-related cloud services are tipped to generate roughly $25 billion in revenue in 2025, sitting against a monstrous $450 billion in infrastructure spend.
Wall Street analysts are already tearing into this massive monetisation gap, and a credible 2029 timeline for commercial quantum injects serious friction into the payback periods for newly minted GPU farms.
Encryption headache
Performance multipliers grab the mainstream headlines, but cryptography remains the immediate, systemic threat.
Analysts at UBS warn that quantum computers will eventually shatter current encryption standards, rendering security nets across finance, healthcare, and sovereign infrastructure utterly useless.
Cyber intelligence outfits have documented a widespread adversarial playbook dubbed harvest now, decrypt later, where bad actors scrape and hoard encrypted data today with the clear intention of cracking it once the hardware catches up.
That global collection phase is already well underway.
China currently dominates global government spending in the sector, with nearly $18 billion invested, pulling aggressively ahead of the EU bloc.
Transitioning to post-quantum cryptographic standards isn't a job data centre operators can kick down the road until the physical hardware actually lands on the racks.
Controlling the transition
Alam's 2029 target isn't a wild punt that quantum will instantly slaughter conventional computing wholesale, as that remains a longer and far murkier horizon.
The near-term reality is a hybrid model, where legacy AI handles the bulk of enterprise workloads while heavy-lift calculations get shunted to quantum processors via the cloud.
Every single major hyperscaler is aggressively laying down the access layers to ensure that handoff is commercially viable.
They have a ruthless financial mandate to ensure computational traffic routes through their own walled gardens rather than leaking to a competitor.
The $600 billion flooding into AI facilities in 2026 is, functionally, the capital base funding quantum's research and development phase.
The mega-caps on both sides of that ledger are the exact same players, laying massive chips across two consecutive eras of computing with the sole intention of monopolising the bridge between them.
