Home/AI Infrastructure/Green Revolution Cooling (GRC)
EN中文
Company

Green Revolution Cooling (GRC)

Austin-born immersion cooling pioneer — the US single-phase immersion incumbent, backed by SK and Samsung as the immersion category re-emerges for AI.

1. Core Product / Service

Green Revolution Cooling (GRC) was founded in 2009 in Austin, Texas — making it the earliest commercial single-phase immersion cooling vendor still standing. Product line as of 2026:

  • ICEraQ — full-tank single-phase immersion for high-density compute (the original product family)
  • ICEraQ SX — newer modular form factor co-engineered with Cisco UCS for AI / HPC
  • ICEraQ Micro — smaller-footprint immersion units for edge and lab deployments
  • Coolant fluids and CDU/auxiliary — designed primarily around SK Enmove and ENEOS dielectric coolants (via investor partnerships)

GRC's pitch is the original immersion thesis: submerge the entire server in a non-conductive dielectric fluid, eliminate fans, drop PUE toward ~1.03, and accept up to ~250 kW/rack densities without the cold-plate-per-chip engineering overhead.

2. Target Users & Pain Points

  • Government / national labs — long-disclosed reference deployments include the NSA and U.S. DoD facilities [6]
  • Telcos — Orange Telecom; Intel has been a co-development customer historically [6]
  • AI / GPU-as-a-Service operators — pursued via the new UNICOM Engineering turnkey partnership announced March 2026 [4]
  • Hyperscalers — limited direct disclosure; immersion's hyperscale penetration remains thin vs DLC

Pain solved: where direct-to-chip cooling requires a chip-specific cold plate per SKU (re-engineering cost every NVIDIA / AMD refresh), immersion is chip-agnostic — you drop the server in a tank and you're done. That theoretical advantage has been the immersion pitch for 15 years; the question has always been adoption.

3. Competitive Landscape

Company Approach Positioning vs GRC
liquidstack Two-phase + single-phase immersion + DLC Trane-acquired (closed 2026); two-phase tech edge
submer Single-phase immersion + immersion AI cloud (InferX) European peer; vertical-integrates into compute
asperitas Single-phase immersion (passive + active convection) Dutch peer; edge / modular focus
coolit-systems DLC pure-play (Ecolab) Different topology — DLC, not immersion
vertiv / schneider-electric / nvent DLC + air Different topology; immersion only as side product

GRC's edge: deepest install base in single-phase immersion in the US + co-investment from SK Enmove and ENEOS (the coolant suppliers, locking GRC into a vertically aligned supply chain) + new Samsung Ventures + Samsung C&T partnership for global scale [1].

4. Unique Observations

  • Pure immersion play — GRC has zero DLC product. This is a one-bet company on the single-phase immersion category winning meaningful AI share. Compare to liquidstack which hedges with both immersion and DLC.
  • NVIDIA / AMD / Intel reference design status: not a public reference partner of NVIDIA's GB200/GB300 stack — that lane is dominated by DLC vendors (coolit-systems, vertiv, Boyd). GRC's chip-vendor relationship is older and historically Intel-aligned (Intel SSD Hub deployments documented). For AI, GRC's wedge is the Cisco engineering alliance on Cisco UCS + ICEraQ SX/Micro and the March 2026 UNICOM Engineering turnkey AI / HPC / GPUaaS partnership [4].
  • Customer breadth: NSA, Orange Telecom, Intel, plus crypto-mining-era customers; AI / neocloud disclosure remains light vs DLC peers. UNICOM partnership is the channel-to-scale bet for 2026.
  • $/MW cooling capex share — immersion claims it can compress total cooling capex via PUE 1.03 (vs ~1.2-1.3 for DLC) and rack densities up to ~250 kW. Of the 15-20% cooling capex in a 1 MW AI build ($3-6M), GRC's stack would replace both DLC cold plates and facility chillers, in theory taking a larger slice (~40-60% of cooling layer, ~$1.5-3.5M/MW). In practice immersion adoption remains in the single-digit-percent of AI deployments vs DLC at the majority of new GB200/GB300 builds — so total addressable revenue today is small.
  • Why immersion lost the first AI round: cold-plate DLC won the NVIDIA reference race because it slots into existing rack/server form factors; immersion requires a tank, new server SKUs (no fans), and operator retraining. The 2026 immersion story is now about second-wave AI — sites that prioritize PUE / water use / sustainability and operators willing to redesign white space.
  • Private with strategic-investor cap table: GRC is not VC-funded in the traditional sense — its lead investors are SK (SK Lubricants / SK Enmove), ENEOS (Japanese oil major), and HTS Group, plus 2025 entry from Samsung Ventures + Samsung C&T [1][2][3]. These are all coolant / industrial / fluid supply chain players — GRC is the front end of an oil-major bet on dielectric fluids as a long-tail consumable in AI DCs.

5. Financials / Funding

  • Status: private; HQ Austin, TX; founded 2009
  • Series C (2024): $28M led by SK Lubricants [2] — brought total funding to ~$43M
  • 2025 round: undisclosed amount from Samsung Ventures; strategic partnership with Samsung C&T; existing investors (HTS, SK Enmove, ENEOS) participated [1][3]
  • Total raised: ~$43M disclosed (pre-2025 Samsung round) [2] — likely higher post-Samsung but amounts not disclosed
  • Revenue / ARR / customer count: not publicly disclosed
  • Valuation: not publicly disclosed

6. People & Relationships

  • Founded: 2009 by Christiaan Best (initial founder)
  • CEO: Peter Poulin
  • HQ: Austin, Texas
  • Investors / strategic partners: SK Lubricants / SK Enmove, ENEOS, HTS Group, Samsung Ventures, Samsung C&T (strategic) [1][2][3]
  • Customers: NSA, Orange Telecom, Intel + government / lab references [6]
  • Engineering alliances: Cisco (Cisco UCS + ICEraQ SX/Micro joint design), LG Electronics (immersion tank co-development for AI DCs, showcased at Data Center World 2026), UNICOM Engineering (turnkey AI / HPC / GPUaaS solutions, March 2026) [4]
  • Strategic context: immersion's "test of the decade" — if AI density keeps escalating past what cold-plate DLC can handle thermally (300+ kW/rack), GRC + submer + liquidstack inherit the next wave. If DLC + rear-door heat exchangers hold the line through GB300, the immersion incumbents stay niche.
Last compiled: 2026-05-11