AI expansion is meeting resistance at both the infrastructure and product layers. OpenAI is buying Pierre Computer to bring code storage deeper into Codex, while voters across the United States approved local limits on data centers. Google is committing another $15 billion to Malaysian computing capacity and tightening school controls around Gemini and Search. At the same time, Sentra and Nooks raised unconventional pools of capital for agent security and sales automation. The shared signal is that AI growth now depends on repositories, data access, power, local consent and institutional controls, not only better models.
OpenAI buys Pierre Computer to expand Codex
OpenAI plans to acquire developer tools startup Pierre Computer and integrate its Code Storage product directly into Codex. The transaction price was not disclosed. Pierre's technology was already used inside Codex, so the deal brings a critical dependency under OpenAI's control as the company expands from code generation toward a broader development platform.
Coding agents create more branches, commits and automated changes than conventional developer workflows. OpenAI introduced its Pando repository backend this week after agent traffic strained existing Git infrastructure. Owning Pierre connects generation, storage and collaboration while making Codex less dependent on outside repository services. The integration still has to preserve familiar Git workflows, permissions and portability for engineering teams.
US voters impose local limits on data center growth
Residents in more than 30 localities across nine states voted for data center moratoriums, bans or other restrictions. Several measures responded directly to projects already under development. The results turn concerns about power prices, water use, noise and land use into binding constraints on the infrastructure behind AI services.
The votes do not end US data center construction, but they make community approval a material deployment risk. Developers can shift projects, yet grid connections and suitable sites are limited. Operators will need clearer evidence on jobs, tax revenue, water use and who pays for new generation and transmission. Public resistance is now a planning variable that investors and model companies cannot treat as a late-stage communications problem.
Google commits another $15 billion to Malaysia
Google said it will invest an additional $15 billion in Malaysia during 2027 and 2028 to expand data center and cloud infrastructure. The company and government project that the program will support about 37,000 jobs. It builds on Google's existing Malaysian data center and cloud-region investments as demand for AI computing grows across Southeast Asia.
Malaysia offers proximity to fast-growing digital markets, but the commitment raises questions about electricity, water and local supply chains. Google paired the announcement with clean-energy and water-replenishment initiatives that will need measurable reporting. The investment shows that AI infrastructure competition is becoming geographically broader even as US projects face stronger political resistance.
Sentra raises $150 million for agent data security
Capital One Ventures led a $150 million investment in Sentra, which maps and restricts the data that human and AI identities can access. A Capital One representative will join the board. Sentra said its business grew more than 300 percent year over year as enterprises deployed copilots and autonomous agents across cloud, software and on-premise data stores.
The product secures data where it already lives instead of inspecting every agent request in real time. It classifies sensitive information, maps inherited permissions and can revoke risky access. Agents often inherit broad rights created for people or service accounts, then use them at machine speed. Funding interest suggests that identity and data governance are becoming prerequisites for enterprise agents.
Nooks gets a $500 million growth commitment
AI sales automation startup Nooks secured a $500 million commitment from General Catalyst's Customer Value Fund. Instead of a traditional equity investment, the fund will receive a share of returns generated by the capital it supplies for sales and marketing. Nooks also launched a Revenue Agent Platform covering workflows from first contact through renewal.
Nooks says it serves more than 2,000 customers and is on track to nearly triple annual recurring revenue this year, although it did not disclose the current figure. The financing structure ties repayment to customer acquisition outcomes and reflects confidence that automation can produce measurable revenue. It also creates execution pressure. Sales agents must improve conversion without damaging customer trust, creating spam or hiding how prospects are scored and contacted.
Google puts school AI behind administrator controls
Google is making Guided Learning available globally in Classroom and piloting the step-by-step tutoring experience in Search. It also said Gemini, Gemini Notebook, Gemini in Classroom and Google Search will be off for K-12 student accounts until school administrators explicitly enable them. The Search pilot begins in selected US schools later this month.
The update places institutional consent ahead of default access while expanding AI features once a school opts in. Guided Learning is designed to coach students through questions rather than immediately supply answers, and Classroom will give teachers real-time signals about where students need help. The policy could reduce unmanaged adoption, but schools still need evidence about learning outcomes, privacy and accessibility. Turning both AI products and Search off by default also makes the administrator decision unusually consequential for basic research access.
What matters today
Today's developments show the AI stack widening in both directions. OpenAI is moving down into repository infrastructure, while Sentra and Nooks are building operational layers above models. Google is expanding physical capacity in Malaysia while accepting stronger gatekeeping in schools. US voters, meanwhile, are asserting direct control over where compute can be built. The next phase of AI competition will reward companies that secure data, finance deployment and earn permission from users, institutions and communities. Capability remains central, but infrastructure legitimacy and operational trust are becoming equally scarce resources.