Theses›Datacentre & AI compute›Datacentre digital twins

Datacentre digital twins

A runtime operational digital twin of the datacentre's compute+networking performance becomes a distinct layer (below design-time EDA, beside the orchestrators) for the non-hyperscaler builders who must model a fleet they cannot afford to over-provision

  • Bear
  • Closing
  • $36B by 2035

The callAI clusters run at 5% GPU use behind a power wall, but NVIDIA gives the twin away; nobody has shown one that predicts a cluster it has not seen.

D1D2R1R2R3C1foragainst D1D2R1R2R3C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the restraints.
  1. D1AI factories cost billions, run against a power wall, and leave GPUs mostly idle, so the builders the giants underserve cannot afford to over-provision.
  2. D2Multi-vendor fleets on AMD and custom silicon arrive, and NVIDIA's free tools only model NVIDIA hardware.
  3. R1NVIDIA gives the gap away: DSX Air launched with a free year, Run:ai, CentML and SchedMD absorbed, Keysight and Spirent inside its tool, Azure building its own.
  4. R2No twin has predicted an unseen cluster without recalibration; AlphaTango claims 1-5% accuracy only on clusters it was trained on.
  5. R3Even the generous runtime slice is about $1B by 2035, so a listed winner is worth well under $1B.
  6. C1AMD's Helios MI450 rack-scale shipments begin, the first multi-gigawatt non-NVIDIA fleet that DSX Air does not model.H2 2026
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028today123predictionswatch 20272028today123predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1AMD Helios fleets in production, and whether independents win customers alongside NVIDIA's tools.Dec 2026 · open
  2. 2By end-2027 no independent runtime datacentre-performance-twin startup reaches a Series A >=$20M with named non-hyperscaler customers paying subscription/seat and a documented cross-cluster moat.Dec 2027 · open
  3. 3An independent runtime twin raising a $20M+ Series A with named paying non-hyperscaler customers and a documented cross-cluster moat.Dec 2027 · open
  • ·A bottom-up size for the runtime compute and networking slice; NVIDIA carries a buy and Cadence a watch, filed elsewhere.
Expressionslisted, private, technologies
  • KEYSKeysight Technologies, Inc.
    • cheap
  • NVDANVIDIA
    • cheap
     Dominant share of AI training compute.
  • SU.PASchneider Electric
    • fair
  • VRTVertiv
    • cheap
  • CDNSCadence Design Systems
    • not screened
     Sells a physics-based (CFD) facility digital twin of data-centre power, cooling and airflow to designers and operators, from its 2022 Future Facilities acquisition.
  • JJacobs Solutions
    • not screened
     Sells a digital-twin service for planning, simulating and optimising gigawatt-scale AI data centres to developers and owners, built on NVIDIA's DSX blueprint.
Private · 4 private companies, 3 of them in the US; none in continental Europe.

AlphaTango · InfrOS Inc. · Phaidra · Red Dot Analytics (RDA)

Technologies
Who has already bet on itprivate rounds, from public sources

$75M raised by 3 companies, $50M of it in the last three years. Wave building: Money is arriving: recent rounds and tier-1 funds are in, but no runaway leader yet.

Backed by a top-tier venture firm: Phaidra. Backed by a strategic investor: Phaidra.