Elon Musk aims to significantly expand the computing capacity of his AI empire, including SpaceX and xAI, to nearly 10 GW by the end of 2027, up from approximately 1.4 GW currently. This ambitious goal comes amidst a surging demand for electricity driven by the AI boom and the proliferation of data centers.

A major hurdle for this expansion is the severe shortage of large gas turbines, which are crucial for continuous power generation. Due to the boom in data centers, major manufacturers like GE Vernova and Siemens Energy have their production slots for these turbines largely booked until 2030.

In response to this supply chain bottleneck, Musk has decided that SpaceX will establish a new foundry in Bastrop, Texas, to manufacture turbine blades and vanes. The goal is to internalize the production of these difficult-to-make components, potentially accelerating the operational readiness of gas turbines by up to 18 months.

Manufacturing turbine blades is an extraordinarily complex process, requiring them to withstand extreme temperatures and centrifugal forces. Experts caution that building a competitive foundry from scratch could take several years, making Musk's projected 18-month saving uncertain.

This situation highlights that in the current AI race, the bottleneck is not just advanced computer chips but also the underlying power infrastructure, specifically the physical components required for electricity generation.

👉 Read the full article on Gizmodo en Español

The OpenAI Agents API provides developers with an interface to build durable cloud agents using a managed Codex harness. This API offloads complex tasks such as session management, orchestration, context compaction, and recovery to OpenAI, while the application supplies tools and chooses its execution environment.

Agents can operate within a sandbox environment, enabling them to execute code, edit files, connect to MCP servers, and produce artifacts. The core concepts of the Agents API revolve around four elements: the "Agent" (comprising the model, instructions, tools, and MCP servers), the "Environment" (an optional sandbox for file access and command execution), the "Session" (a durable instance of an agent working on tasks), and "Events and items" (the inputs and outputs of a session).

The managed harness offers a wide range of capabilities, including running commands in a sandbox, applying relevant skills and instructions, connecting to external data via tools or MCP, steering the agent during its work, summarizing past work for context window management, delegating subtasks to subagents, and resuming sessions.

Pricing for the Agents API is based on the selected model's API rates, standard rates for OpenAI tools, and standard container rates for OpenAI-hosted sandboxes. Currently, the Agents API supports data residency only in the United States and does not offer Zero Data Retention (ZDR).

Example applications include creating and running directory-tree scripts, comparing and combining release notes using subagents, incident response agents, Slack bots, data analysts, GitHub issue investigators, and document reviewers.

👉 Read the full article on OpenAI Developers

According to TechCrunch, OpenAI has temporarily paused new sign-ups for its Pro subscriptions due to overwhelming demand for its new GPT-6 Astra model. This measure is likely intended to maintain service quality for existing users and manage the rapidly increasing resource requirements.

The GPT-6 Astra model, with its advanced capabilities and performance, has garnered significant interest since its release, placing substantial strain on OpenAI's infrastructure. The company is presumably focusing on scaling its capacity to meet this surging demand.

This subscription pause highlights the growth challenges OpenAI is facing, underscoring the unexpected surge in demand for cutting-edge AI models upon their market introduction and the subsequent infrastructure hurdles. Users will need to await the resumption of Pro subscriptions.

👉 Read the full article on TechCrunch