LiquidStack

Bitfury-spun two-phase immersion specialist now folded into Trane Technologies — the most technically distinct cooling company in the M&A wave.

1. Core Product / Service

LiquidStack started inside the cryptocurrency miner Bitfury Group (acquired by Bitfury in 2015) and spun out as an independent company in 2021 [6]. Its hybrid product line straddles the immersion / DLC boundary:

• Two-phase immersion — the original hallmark; servers submerged in a low-boiling-point dielectric that phase-changes at chip surface for vastly higher heat flux than single-phase immersion (grc-cooling, submer, asperitas are all single-phase)
• Single-phase immersion — added later for broader compatibility
• Direct-to-chip (DLC) cold plates — added to the portfolio to compete with coolit-systems / vertiv on conventional AI BoMs
• Coolant Distribution Units (CDUs) — facility-side liquid management

Trane Technologies signed definitive agreement to acquire LiquidStack (announced Q4 2025; expected close early 2026) [1][3]. Trane first invested as part of the Series B in March 2023 alongside Tiger Global [1]. Status: acquired by Trane / pending integration into Trane's thermal management business.

2. Target Users & Pain Points

• Hyperscalers — NTT (Japan) is a publicly-disclosed two-phase immersion customer; Standard Power (US neocloud) also deploys LiquidStack
• HPC operators — high-density compute facilities needing >100 kW/rack
• Crypto mining (legacy) — original Bitfury-era use case; less relevant for the AI thesis
• Chip OEMs — Wiwynn (the lead in LiquidStack's $10M Series A in 2021) is a strategic OEM partner [6]

Pain solved: at >200 kW/rack — the regime GB300+ and post-NVL72 architectures approach — single-phase immersion and even DLC start hitting limits. Two-phase immersion uses latent heat of vaporization to absorb 10-100× more thermal energy per liter of coolant than single-phase, in theory giving headroom for the next two generations of GPU.

3. Competitive Landscape

LiquidStack's edge under Trane: it is the only major company in the cooling M&A wave that brings two-phase immersion IP. Trane gets a hedge against the DLC-dominant world that coolit-systems / Boyd / motivair are building — a play on the GPU density curve overshooting DLC's thermal envelope.

4. Unique Observations

• Hybrid technology bet — Immersion + DLC: LiquidStack is the rare cooling company that straddles both architectures. The headline IP is two-phase immersion (high-flux, phase-change dielectrics), but the company shipped a DLC product line to participate in the dominant short-term AI demand. This makes it the most technologically optionality-rich pick among immersion peers.
• Two-phase is the unique technical asset: most immersion peers are single-phase only. Two-phase systems handle higher density and recover heat at usable temperatures more efficiently — but require more sophisticated fluid management and have historically faced regulatory headwinds (some early dielectrics like 3M's Novec were phased out for PFAS concerns). LiquidStack has had to navigate fluid chemistry transitions in parallel with growth.
• NVIDIA / AMD / Intel reference design: not a public NVIDIA reference partner on the GB200 cold-plate side — that lane belongs to coolit-systems, vertiv, and Boyd. LiquidStack's NVIDIA-relevant story is via OEM partner Wiwynn (a key NVIDIA HGX-system maker, who led the 2021 Series A) [6].
• Customers (publicly disclosed): NTT (Japan), Standard Power (US neocloud) — plus quiet hyperscaler trials. Customer count is small relative to DLC peers but average deal size large.
• $/MW cooling share: at two-phase immersion's density (250 kW/rack and up), the technology can compress facility-side capex (no chillers, lower-temperature recovery loops). Of the 15-20% cooling capex layer in a 1 MW AI build ($3-6M), a full two-phase deployment can capture 40-50% of cooling capex (vs ~25-35% for DLC cold plates alone). But penetration today remains in low single digits of new AI capacity; total served-addressable revenue is small.
• Trane acquisition logic: Trane brings global HVAC scale, chiller manufacturing, and hyperscale customer relationships. LiquidStack brings the two-phase IP and the DLC product line. Together it parallels what eaton + Boyd did at the larger scale (with DLC, not immersion). The price wasn't disclosed, which is itself notable — Trane is private about the multiple, in contrast to Eaton's $9.5B / coolit-systems's $4.75B headlines.
• Funding history: $10M Series A (2021, Wiwynn-led) → Series B March 2023 with Trane + Tiger Global → Trane definitive agreement Q4 2025 / close 2026 [1][3][4][6]. Disclosed total raised: not public; estimates ~$50-80M range pre-acquisition.

5. Financials / Funding

• Status: definitive agreement to be acquired by Trane Technologies; expected close early 2026 [1][3]
• Series A (2021): $10M, led by Wiwynn [6]
• Series B (March 2023): Trane Technologies + Tiger Global (amount not disclosed) [1][4]
• Trane acquisition (2026): deal value not publicly disclosed — unusual relative to peer deals
• Revenue / ARR: not publicly disclosed
• Valuation: not publicly disclosed

6. People & Relationships

• Founded: spun out from Bitfury Group (which acquired the predecessor business in 2015); independent company since 2021 [6]
• CEO: Joe Capes (CEO since spin-out)
• HQ: Manchester, NH (US) — historically also Hong Kong (Bitfury heritage)
• Investors / lineage: Bitfury (parent until 2021) → Wiwynn (Series A lead) → Trane Technologies + Tiger Global (Series B) → Trane (acquirer 2026) [1][4][6]
• Customers (public): NTT, Standard Power
• Strategic context: with Trane absorbing LiquidStack, every major Tier-1 cooling specialist has now found an industrial parent (Boyd→Eaton, CoolIT→Ecolab, Motivair→Schneider, Chilldyne→Daikin, JetCool→Flex, LiquidStack→Trane). The independent pure-play era of the AI cooling cycle effectively closed in 2025-2026.