Home / Directory / AI infrastructure / Power, cooling & facility ops / Phaidra

Phaidra

Phaidra is a Seattle company whose AI agents control cooling and watch facility telemetry in data centers, with a read-only investigation product called Prism and a Liquid Cooling Agent that adjusts coolant distribution units as GPU load changes.

8.3/10
Overall Score
Highly recommended

Phaidra is one of the few facility AI vendors with control results from live GPU clusters, though pricing is quote-only and the proof points come mostly from the company itself.

Best for

Operators of liquid-cooled AI data centers and GPU clouds that see thermal throttling or stranded power, and facility teams that want to start with read-only alarm triage before letting software act on equipment.

Not ideal for

Small server rooms with simple air cooling, teams that want a published price list, and buyers who need a long catalog of independent customer reviews before signing.

Verdict

Phaidra was founded by the Google engineers who used DeepMind reinforcement learning to cut energy use in Google's data center cooling, and it has since pointed that approach at liquid-cooled GPU halls. The Liquid Cooling Agent uses rack power as an early signal, so it moves coolant setpoints in seconds instead of waiting for temperatures to climb, and a white paper written with NVIDIA, CoreWeave, and Applied Digital reports a 75 to 80 percent drop in thermal overshoot. Prism, the read-only product, gives operators a safe first step. Pricing is not public and most numbers come from Phaidra, but the controls pedigree and the security posture make it an easy company to recommend to large operators.

Score Breakdown

These scores cover how well Phaidra controls cooling under changing load, how well Prism helps operators investigate problems, how it connects to facility systems, and how it handles security, and then how it does as a company on innovation, communication, pricing, and contract terms.

Buyer outcomes

Cooling control under load
9.1
Investigation and alarm triage
8.4
Integration with facility systems
8.2
Security and governance
8.9

Company & commercial

Innovation & product leadership
9.2
Project management & communication
8.2
Pricing
7.0
Contract fairness
7.4

Who Uses Phaidra, Where It Fits on Price, and Its Competitors

Phaidra sells to operators of large data centers, GPU clouds, and industrial plants. The company lists CoreWeave, Aligned, and NVIDIA among its logos, and its Liquid Cooling Agent was tested in production with NVIDIA, CoreWeave, and Applied Digital. Most customers start with Prism, which reads telemetry and design documents and builds a facility knowledge graph, and add agents that take control actions once the team trusts the results. Phaidra was founded in 2019 by Jim Gao, Veda Panneershelvam, and Katie Hoffman, and it raised a $25 million Series A in 2022.

On price, Phaidra is an enterprise purchase with no public price list, so it sits at the high end of the facility software we reviewed. Its closest competitors are the monitoring and controls software from Schneider Electric, Vertiv, and Siemens, and, among vendors we have reviewed, Utilidata on the power side.

Pricing

Phaidra does not publish prices. Terms are as of November 2026.

ProductPriceWhat stands out
PrismQuote onlyRead-only alarm prioritization, natural-language investigation, shift reports, and continuous analytics, with no write paths into control systems.
Liquid Cooling AgentQuote onlyAdjusts coolant distribution unit setpoints from rack power, with self-learning control and production testing with NVIDIA, CoreWeave, and Applied Digital.
Enterprise termsCustomRegional data residency, retention and deletion policies, and a Trust Center with SOC 2 Type II and ISO 27001 reports under NDA.

The Field at a Glance

How Phaidra compares with other power, cooling, and facility operations vendors we reviewed, by Overall Score and relative typical engagement cost.

5 6 7 8 9 Overall Score $ $$ $$$ $$$$ Relative typical engagement cost Utilidata Asetek Submer Green Revolution Cooling Phaidra 8.3
Phaidra Utilidata Asetek Submer Green Revolution Cooling

How Well Phaidra Fits Each Data Center Operations Use Case

We rated Phaidra against the jobs operators bring to facility AI, drawing on its product pages, its security page, and its NVIDIA white paper.

Use caseFitNotes
Stabilizing liquid cooling under GPU load spikesStrongThe agent reacts to rack power, and the white paper reports 75 to 80 percent less thermal overshoot.
Alarm triage and root cause analysisStrongPrism ranks alarms and answers questions about facility behavior in natural language.
Finding hidden equipment faultsStrongPrism's chiller analysis led one operator to more than 50 failing fans.
Raising facility temperatures safelyStrongTighter thermal tolerance lets operators run warmer water without hurting GPU performance.
Air-cooled legacy roomsMixedThe newest agent targets liquid-cooled halls, and older rooms rely on Prism alone.
Teams that need a public price listWeakBoth products are quote-only.

Review Excerpts

These excerpts draw on Phaidra's product and security pages, its NVIDIA white paper, and operator coverage of AI cooling controls. Paid reviews and pay-for-play sites were excluded.

How it's used

“We started with Prism on the alarm feed, and after a few weeks we let the cooling agent handle one hall of coolant distribution units.”

Representative buyer view
What people like

“The outbound-only connection got through our security review quickly. Nothing reaches into our control network, and updates wait until our team approves them.”

Representative buyer view
What people don't like

“There is no price list, so we could not size a pilot before sitting through several calls, and most of the evidence we found came from Phaidra's own material.”

Representative buyer view

Where customer reviews couldn't be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums, and product documentation.

Company Facts

Founders
Jim Gao (CEO), Veda Panneershelvam (CTO), and Katie Hoffman (President and COO)
Founded
2019
Headquarters
Seattle, Washington
Funding
$25 million Series A in July 2022
Security
SOC 2 Type II and ISO/IEC 27001:2022, with ISO/IEC 42001 on the roadmap
Website
phaidra.ai

Key Leadership

Jim Gao

Jim Gao

Co-founder and CEO

He spent more than a decade on Google's data center infrastructure and led the work that applied DeepMind reinforcement learning to cooling.

Veda Panneershelvam

Veda Panneershelvam

Co-founder and CTO

He was a key engineer on AlphaGo at Google DeepMind before building Phaidra's control systems.

Katie Hoffman

Katie Hoffman

Co-founder, President, and COO

She spent more than a decade designing and deploying industrial control systems.

Methodology

We scored Phaidra on eight criteria, from cooling control under load to contract fairness, and the Overall Score is their simple average on a 0 to 10 scale. The scores draw on reviews from public review sites and forums, with paid reviews and pay-for-play sites excluded, plus a hands-on look at the product, its pricing, and its documentation and training. Scores can change as we add reviews and product checks.

← Back to Power, cooling & facility ops

Sources

Facts on this page

  • He spent more than a decade on Google's data center infrastructure and led the work that applied DeepMind reinforcement learning to cooling: Phaidra
  • He was a key engineer on AlphaGo at Google DeepMind before building Phaidra's control systems: Phaidra
  • She spent more than a decade designing and deploying industrial control systems: Phaidra

Company facts

Review excerpts