Fluidstack's 197 Open Roles Map Anthropic's $50B Gigawatt Buildout
Fluidstack doubled open roles to 197 in two months to deliver Anthropic's $50B gigawatt buildout. Here is the four-city sourcing map recruiters need.
Fluidstack's July 1, 2026 "Who is hiring?" post on Hacker News is the first public, city-by-city map of who is actually staffing Anthropic's $50 billion compute buildout. It names SFO, SEA, NYC, and ATX as onsite hubs, then quietly lists SWE, SRE, Infrastructure, Security, Product, and Design across all four. Most sourcers will read that and open LinkedIn. That is the wrong move, and this post explains why.
The deal, in numbers you can actually use
Anthropic announced the $50 billion partnership with UK-based neocloud Fluidstack on November 12, 2025. The first custom data centers land in Texas and New York, coming online throughout 2026. Public reporting pegs the buildout at 800 permanent jobs and more than 2,000 construction roles. Fluidstack's own careers page frames the ambition as "operate at the scale of a nation, not a building" and calls out 100 GW by the end of the decade, roughly the entire electricity consumption of Japan.
The hiring signal is already visible on the ground. ZipRecruiter lists 197 open Fluidstack jobs as of July 2026, up from roughly 97 in May. That is a doubling in two months, timed to a company that just repriced from a $7.5B valuation to $18B in April 2026 on a round led by Jane Street and Leopold Aschenbrenner's Situational Awareness fund.
The demand pull is not speculative. Anthropic's run-rate revenue crossed $30 billion in 2026, up from about $9 billion at the end of 2025. That is why Fluidstack's headcount plan has an accelerator pinned to the floor.
The four-city framing is a trap
Here is where recruiters lose. SFO, SEA, NYC, and ATX are HQ and hub cities for software, SRE, product, and design. The actual gigawatt is not being built in any of them.
The Texas capacity is Fluidstack's 168MW Abernathy facility, co-developed with TeraWulf, plus a 244MW project with Cipher Mining. Abernathy is a town of about 2,800 people north of Lubbock. It is not Austin. The New York capacity is a 360MW Fluidstack-TeraWulf project at Lake Mariner, in upstate NY near Barker. It is not Manhattan. Hut 8 has a separate 245MW deal tied to the same partnership.
If you source facility operators, high-voltage electricians, chiller techs, or OSP fiber engineers using an "NYC" or "Austin" geo filter, you will find zero of the people who will actually run these sites. The onsite requirement for those roles is a rural exurb next to a substation.
What each city hub actually needs
- SFO: Product, Design, senior SWE for the control plane, BMS/DCIM tooling PMs. Culture pull toward the Anthropic orbit.
- SEA: Infrastructure SWE, storage, networking software, likely leveraging the AWS/Azure alumni base.
- NYC: Business, finance, capital markets adjacencies (Jane Street is a lead investor, that is not an accident), plus some SWE.
- ATX: Bridge to the Texas builds. SRE, network engineering, integrator and supplier quality managers who can drive to Abernathy.
For the actual datacenter shells, you are recruiting into TeraWulf's Lake Mariner, Cipher's West Texas sites, and Hut 8's operating footprint. Different pool. Different pitch.
The scarce skill is not GPUs
Every recruiter with a pulse is chasing "GPU engineers" and "InfiniBand" this year. Fluidstack's job postings tell a different story. The scarce, urgent skills are:
- High-voltage substation and utility interconnect work (ERCOT and NYISO experience is gold)
- Liquid cooling, CDU, and rear-door heat exchanger operations
- UPS, PDU, CRAC, and chiller literacy at megawatt scale
- OSP (outside plant) fiber, OTDR troubleshooting, diverse path planning, peering and interconnection
- Multi-vendor NOS: Arista, Juniper, SONiC, Cisco CLI in the same posting
That last bullet is the tell. Fluidstack is explicitly asking for engineers fluent across Arista, Juniper, SONiC, and Cisco. That combination does not live at FAANG. It lives at hyperscalers with white-box fabrics, at telcos, and increasingly at Bitcoin miners who pivoted to AI.
The scarce skill is not GPU expertise. It is high-voltage substation, liquid cooling, and OSP fiber, and it does not live at FAANG.
The Bitcoin-miner-to-AI-neocloud lane
The single most productive sourcing lane for Fluidstack hiring in 2026 is the Bitcoin miner alumni pool. TeraWulf, Hut 8, Cipher Mining, Core Scientific, Applied Digital, and Crusoe all have the exact skill mix Fluidstack needs: megawatt-scale power procurement, remote-site operations, rapid buildout, and increasingly, GPU cluster operations.
Paul Prager runs TeraWulf. Asher Genoot runs Hut 8. Their operations staff already know how to negotiate with ERCOT, run substations, and pour concrete on aggressive timelines. Half of them have been quietly retraining on HGX platforms for two years.
The problem is that these people are hard to find with keyword search. Their LinkedIn titles say "Site Operations Manager" or "Facility Engineer" at companies that most tech recruiters have never sourced. Their GitHub is thin because their day job is power and pipe, not pull requests. This is exactly the shape of query that boolean strings fail on and plain-English search wins on, which is why we built Refolk: describe "facility engineers at Bitcoin miners who now operate GPU clusters in West Texas or upstate New York" and get a ranked shortlist, not a 4,000-row export.
Why the LinkedIn pool is deceptive
Refolk's index shows roughly 39,000 US-based profiles with "Data Center Engineer" or "Network Engineer" as their title. That looks like abundance. It is not.
Filter that pool for people who have actually operated RDMA fabrics at scale, and it collapses fast. Top current employers in the raw pool skew toward J.P. Morgan, municipal governments, and enterprise integrators. Nothing wrong with those engineers, but they are not who Fluidstack needs shipping a 168MW Abernathy site by year-end.
The AI-native operators (Fluidstack itself, CoreWeave, Lambda, Nebius, Crusoe) hold a small, tightly held population. CoreWeave alone is working through a $21B Meta contract and total committed contracts exceeding $35B. They are bidding for the same people. So are the Stargate sites in West Texas, Anthropic's own $30B Azure/NVIDIA capacity, and the AWS Rainier team building the up-to-5-gigawatt Amazon agreement.
Roles most sourcers will miss entirely
If you only read the HN post's headline, you will chase SWE and SRE. Fluidstack's actual posting board is broader. From Builtin and the fluidstack.io careers page:
- Connectivity Engineer. Manages carrier relationships, procures and tracks circuits, ensures diverse OSP fiber paths, coordinates fiber builds, plans backbone and POP placement, supports peering and interconnection. This is a telco skill, not a cloud skill.
- Integrator Quality Manager and Supplier Quality Engineer. These are manufacturing-adjacent roles. Source from Tesla Energy, Applied Materials, and semiconductor fab supply chains.
- Community Engagement Lead. For public hearings and permitting fights around new datacenter sites. Most tech recruiters will not even see this as a tech role. It is the difference between a 168MW site coming online on time and slipping six months.
- Data Center Facility Operator and Tier 3 Operations Technician. Base band published at $100,000 to $150,000 plus equity, depending on experience and location.
The Community Engagement Lead is worth calling out separately. Every gigawatt buildout in 2026 is running into local opposition over water, power, and noise. Fluidstack is hiring for that fight explicitly. That role wants former utility PR, campaign staff, or municipal government affairs, not a bootcamp SWE.
The pitch that works on this pool
Fluidstack's culture text is unusually explicit: "Extreme ownership. Full autonomy. Own things end to end often taking on scope outside your core role without being asked to get things done. Velocity. We drive everything forward as fast as possible." Gary Wu, the co-founder and CEO, told Anthropic press that "Fluidstack was built for this moment."
Two implications for outbound:
- Downweight FAANG pedigree filters. The Situational Awareness investor signal pulls in AGI-pilled ICs who want mission narrative more than TC ladder. Ex-miner ops engineers and ex-telco fiber engineers fit the culture better than ex-Meta L6s. This is a common failure mode for GPU networking recruiters who default to the same 200 LinkedIn profiles everyone else pings.
- Lead with the buildout, not the perks. "168MW in Abernathy, 360MW at Lake Mariner, gigawatt by end of 2026, Anthropic anchor" beats a comp band in a first message every time for this audience.
For AI infrastructure sourcing at this speed, the workflow that scales is plain-English description feeding a ranked shortlist, then human outreach. Refolk pulls across GitHub, LinkedIn, and the open web in one query, which matters here because the fiber and facility population barely shows up on GitHub, and the software population barely shows up on datacenter conference attendee lists.
A 30-day sourcing plan for Fluidstack hiring
- Week 1. Map every named counterparty: TeraWulf, Cipher Mining, Hut 8, Core Scientific, Applied Digital, Crusoe. Pull site operations, facility engineering, and network engineering rosters. Filter for people already within 90 minutes of Abernathy, Lake Mariner, or Cipher's West Texas sites.
- Week 2. ERCOT and NYISO operator alumni. Utility-side power engineers on LinkedIn who have posted about interconnect queues in the last 18 months. This is the substation talent.
- Week 3. Multi-vendor NOS engineers. Boolean will fail. Describe the profile in plain English (Arista and Juniper and SONiC in the same resume, plus OTDR) and let a semantic search do the work.
- Week 4. The wildcards. Community Engagement Leads from utility PR and municipal affairs. Integrator Quality Managers from Tesla Energy and semiconductor supply chains. These close fast because nobody else is calling them.
Fluidstack hiring is going to accelerate through Q4 2026 as the Abernathy and Lake Mariner sites approach commissioning. The recruiters who win this cycle will have built their shortlist against the actual facility footprint, not against the four HQ cities in the HN post.
FAQ
Where are Fluidstack's Anthropic-funded data centers actually located?
The Texas capacity is centered on the 168MW Abernathy facility (co-developed with TeraWulf, north of Lubbock) and a 244MW Cipher Mining site. The New York capacity is a 360MW Fluidstack-TeraWulf project at Lake Mariner in upstate NY. Hut 8 has a separate 245MW deal tied to the same partnership. Despite Fluidstack's HN post listing SFO, SEA, NYC, and ATX as onsite hubs, none of the actual datacenter shells are in those cities.
Who is Fluidstack competing with for the same engineers?
CoreWeave (working a $21B Meta contract with more than $35B in total committed contracts), Lambda, Nebius, and Crusoe are the direct AI-neocloud competitors. Beyond that, Anthropic's own $30B Azure and NVIDIA agreement, the up-to-5-gigawatt AWS Rainier partnership, the 5-gigawatt Google and Broadcom partnership starting in 2027, and Stargate's West Texas sites are bidding for overlapping talent pools.
What comp should recruiters expect to quote?
Fluidstack has published a base band of $100,000 to $150,000 plus equity for a Data Center Operations Technician role, depending on experience, skills, qualifications, and location. Software, SRE, and infrastructure roles at the SFO and SEA hubs will sit meaningfully higher. Equity is a real component given the company's April 2026 repricing from $7.5B to $18B.
Which sourcing lane is most productive right now?
Bitcoin-miner-to-AI-neocloud alumni. TeraWulf, Hut 8, Cipher Mining, Core Scientific, Applied Digital, and Crusoe engineers already own the megawatt power, remote-site operations, and rapid buildout skill mix Fluidstack needs. They are hard to surface through keyword search because their titles and GitHub profiles do not scream "AI," which is why plain-English semantic search across LinkedIn and the open web outperforms boolean strings on this specific pool.