Aller au contenu
Toutes les missions
Veille marché · Dev-to Publiée il y a 2 semaines

I read all 308 open jobs at xAI: 54% are in its data center team, and most of them are in Memphis

A job board is a company's plan written as headcount. On 31 October 2026 I pulled every open posting from xAI's public Greenhouse board and counted them by the department label xAI itself as...

Entreprise xAI
Mode g from xAI

Résumé

A job board is a company's plan written as headcount. On 31 October 2026 I pulled every open posting from xAI's public Greenhouse board and counted them by the department label xAI itself assigns. I expected model researchers. I found construction superintendents. Disclosure: I built the data tool used here. Details and a way to reproduce this are at the end. The headline number xAI had 308 open jobs. 167 of them (54%) are filed under the department "Data Center". The next biggest departments are far smaller: Department (xAI's own label) Open jobs Data Center 167 Engineering 26 Human Data 14 Information Security 13 Product 13 Safety 12 Sales 11 Operations 10 Model 7 Legal 7 Other (9 small departments) 28 Total 308 Only 7 jobs sit in the department called "Model". Post-Training (3) and Vision (2) are even smaller, and those five are all "AI Tutor" postings. The board is not hiring a research army. It is hiring the people who build and run the building the research runs in. One small oddity: the board lives at job-boards.greenhouse.io/xai, but the company name it reports on each posting is "SpaceXSI". I simply report what the board says. What "Data Center" means in practice I searched the 167 Data Center titles for a few keywords (a title can match more than one): Keyword in title Jobs engineer 50 technician 36 construction 28 electric (electrical, electrician) 26 manager 22 mechanical 11 superintendent 10 fiber 9 controls 6 commissioning 6 power plant 5 Some real titles from the board, verbatim: Construction Superintendent, Power Generation - Memphis Director, Power Plant Operations - Memphis Asset Manager, Energy Storage - Memphis Chiller Technician - Memphis Certified Weld Inspector - Memphis Construction Rigger - Memphis Fire Protection Engineer - Memphis HVAC Supervisor, Chilled Water Systems - Memphis Construction Superintendent, Electrical - Night Shift (Memphis) Power generation, battery storage, chilled water and night shifts. That is the vocabulary of a utility or a large construction site, not of a software company. Where the jobs are Almost all Data Center roles point to one area: Location (as written on the posting) Data Center jobs Southaven, MS; Memphis, TN 132 Memphis, TN 30 Southaven, MS 3 Other 2 That is 165 of 167 in the Memphis and Southaven area, right on the Tennessee and Mississippi border. Across the whole board, the top locations were Southaven/Memphis (174 postings) and Palo Alto (74). The 22 "AI Tutor" jobs The board lists 22 "AI Tutor" roles (12 in "Human Data", the rest in STEM, Post-Training, Vision and Legal & Compliance). All 22 are marked temporary, and many are language-specific (examples: "AI Tutor, Uzbek - Temporary", "AI Tutor, Oromo - Temporary", "AI Tutor, Latvian - Temporary"). These are people who write and grade training data in specific languages. It is a small but telling line item: frontier models still need humans for the long tail of languages. How this compares to two software-first AI companies Same day, same tool, title-based counts (a rougher method than department labels, so treat these as approximate): Company Open jobs Engineering titles Customer-facing titles* OpenAI 252 88 34 Perplexity 138 65 22 Cursor 125 42 59 xAI 308 118 12 *Sales, solutions, success, partnerships, marketing, business development, forward-deployed and similar. xAI posts the fewest customer-facing roles of the four and by far the most physical-infrastructure roles. Cursor sits at the other extreme: more customer-facing titles than engineering titles. What I take from it Compute is now a construction business. When more than half of an AI lab's open roles are about power, cooling, fiber and concrete, the bottleneck is physical, not algorithmic. Location follows power. The jobs cluster where the site and its power generation are, not where AI talent traditionally lives. Job boards are an early signal. Hiring for a power plant director and night-shift construction superintendents tells you about build-out plans months before a press release does. This is one snapshot. Job boards change daily, and a posting is a plan, not a hire. How to reproduce this I pulled the board with my Career Site Jobs API on Apify. It reads public Greenhouse, Lever, Ashby, Workday and similar boards and returns one clean JSON row per job (title, department, location, salary range if published, URL). Input I used: { "companies": ["greenhouse:xai", "ashby:openai", "ashby:perplexity", "ashby:cursor"], "includeDescription": false, "maxJobsPerCompany": 2000 } Pricing is $0.002 per job returned, so the 308 xAI jobs cost about $0.62. Counting by department is then a few lines of Python: from collections import Counter jobs = [j for j in rows if j["company"] == "SpaceXSI"] print(Counter(j["department"] for j in jobs).most_common()) If you track a company's build-out, run it on a schedule and diff the results: new departments and new locations show up there first.