Tech Current Daily Brief — September 2, 2026

The strongest technology signal from the last 24 hours is that AI is reshaping institutions as much as products. Apple formally began its first CEO transition since 2011, putting longtime hardware chief John Ternus in charge as the company faces an AI-centered product cycle. SB Energy filed for a U.S. IPO that could become one of the clearest public-market tests of the AI data-center boom. OpenAI confirmed that Astra is its first model to cross the company’s Critical cybersecurity threshold, forcing a more restrictive release strategy. Autonomous mobility kept scaling as Waymo and Zoox expanded into more U.S. cities. And a new security category is forming around the tools agents install and invoke, with AIR emerging from stealth with $50 million to police skills, plug-ins and MCP servers.

Today’s 5 Top Stories

Apple begins the John Ternus era as Tim Cook becomes executive chairman

John Ternus formally became Apple’s chief executive on September 1, replacing Tim Cook after a 15-year run that transformed Apple into one of the world’s most valuable companies. Apple had announced the succession in April, but the handoff became effective this week: Ternus joined the board and Cook moved into the role of executive chairman, where Apple says he will continue to assist with areas including engagement with policymakers. Ternus has spent roughly 25 years at Apple and most recently led hardware engineering across products including iPhone, iPad, Mac, Apple Watch and AirPods.

Why it matters: This is not simply a management change. Apple is entering a period in which its device advantage, custom silicon strategy and privacy model all have to translate into credible AI products. A hardware engineer taking the top job suggests Apple may continue to treat on-device compute and tightly integrated product design as its primary route into the AI era rather than copying cloud-first rivals.

SB Energy files for a U.S. IPO as AI infrastructure moves toward the public markets

SB Energy publicly filed an S-1 on September 1 for a proposed Nasdaq and Nasdaq Texas listing under the ticker SBE. The SoftBank-backed company is positioning itself as a vertically integrated power-and-data-center platform for the AI era, with OpenAI and SoftBank already tied to major campuses and Nvidia expected to become an investor in connection with the offering. The filing shows 8.8 gigawatts of data-center capacity contracted or under construction, while Reuters reporting tied to the filing says the company has roughly $439 billion of backlog and could seek a valuation above $50 billion. The risk is equally striking: SB Energy currently has no operational data centers and remains highly dependent on turning long-duration AI contracts into functioning infrastructure and cash flow.

Why it matters: The AI infrastructure boom has mostly been financed through hyperscaler capex, private equity, debt and private-company balance sheets. SB Energy’s IPO would force public investors to price the power, construction, customer-concentration and execution risk behind gigawatt-scale AI campuses. It is a direct test of whether the market will value contracted AI demand before the underlying data centers are operating.

OpenAI confirms Astra is its first model with Critical cybersecurity capability

OpenAI said on September 1 that Astra now meets the Critical cybersecurity capability threshold under its Preparedness Framework. The company says Astra can, with the right tools and access, find previously unknown vulnerabilities and develop exploits across well-protected systems without a person guiding every step. In one internal benchmark it achieved a 100% score on ExploitBench, and during a separate evaluation it discovered and used two previously unknown vulnerabilities in an exploit chain. OpenAI says it delayed parts of Astra’s development and release while strengthening safeguards and plans to make the model available soon, but its most advanced cyber capabilities will initially be restricted to a smaller group of testers and defensive-access programs.

Why it matters: Frontier AI has crossed an important operational threshold: zero-day discovery and exploit development are becoming model capabilities that labs expect to deploy, not hypothetical future risks. The competitive question is no longer only who has the smartest model, but who can release systems with offensive-grade capabilities without making high-end cyber exploitation dramatically cheaper for attackers.

Waymo and Zoox widen the U.S. robotaxi race with new-city expansion

Alphabet’s Waymo and Amazon’s Zoox both announced new U.S. expansion on September 1. Waymo began welcoming its first public riders in Denver, San Diego and Tampa, bringing its fully autonomous public-ride footprint to 14 cities and introducing more users to its next-generation vehicle platform. Zoox said it would begin testing in Houston and San Diego, taking its presence to 12 U.S. locations after launching paid rides in Las Vegas in August. Zoox will start with manually driven mapping and test vehicles before deploying its purpose-built driverless robotaxis.

Why it matters: Robotaxis are moving from city-by-city demonstrations into a repeatable expansion problem. The differentiator is increasingly operational: fleet manufacturing, mapping, local approval, remote assistance, maintenance and the speed at which a company can replicate a driverless service across markets. Waymo’s public-rider expansion and Zoox’s simultaneous geographic buildout show that autonomy is becoming a scaling contest rather than only a technology contest.

AIR raises $50 million to secure the software supply chain around AI agents

AI security startup AIR emerged from stealth on September 1 with $50 million raised across two seed rounds led by Sequoia and Greenoaks. Its product is designed to discover agents running inside enterprises, monitor the skills, plug-ins, MCP servers and other components those agents use, and block tools or external sources that fail security criteria. AIR says it continuously re-checks agent add-ons because a previously safe package can become dangerous after a dependency changes or an account is compromised. The company says it currently filters out roughly 27% of the add-ons and skills it finds online and already has more than 20 customers, including large enterprises.

Opportunity signal: AI agents are creating a new software supply chain before the security model around it is mature. Skills and MCP servers increasingly function like executable extensions with access to credentials, data and enterprise systems. That creates room for a new security layer built around provenance, continuous verification and runtime policy—similar to how operating systems eventually standardized signed drivers and application permissions.

Data & Market Pulse

Today’s numbers show three different forms of AI scale. SB Energy is presenting 8.8 gigawatts of contracted or under-construction data-center capacity to public-market investors. OpenAI’s Astra reached a 100% score on its ExploitBench evaluation while crossing the company’s Critical cyber threshold. AIR says its security pipeline rejects about 27% of the agent skills and add-ons it encounters online. At the same time, Waymo has reached 14 cities with public fully autonomous rides, while Zoox is now present in 12 U.S. locations. Capital, capability and deployment are all scaling faster than the control systems around them.

Trend Watch

1. AI infrastructure is becoming a public-market asset class. SB Energy’s filing gives investors a chance to price not just chips or cloud revenue, but long-duration power, construction and lease commitments tied directly to frontier-model demand.

2. Cyber capability is forcing differentiated model access. OpenAI’s Astra strategy points toward a future in which the most capable frontier models are not released with a single universal permission set; access may vary by capability, user trust and defensive use case.

3. Agent ecosystems are recreating old software-security problems at higher speed. As agents load skills, call external servers and take actions autonomously, provenance and runtime control are becoming as important as model alignment itself.

What to Watch Next

For Apple, the first major product cycle under Ternus will show whether leadership continuity translates into a more aggressive AI product strategy. For SB Energy, watch for the IPO price range, final Nvidia participation and how investors value a company whose largest AI data-center assets are still being built.

For OpenAI, Astra’s system card and final access rules will be the clearest test of how frontier labs intend to commercialize critical cyber capabilities. For autonomous mobility, watch how quickly Waymo converts its new-city waitlists into broad public access and when Zoox moves from mapping to purpose-built vehicles in Houston and San Diego. In agent security, the important signal is whether enterprises start requiring signed or continuously verified skills and MCP servers as a standard procurement requirement.