The AI Superapp Race Is On, EVs Are Pivoting, and Builders Are Betting on Power

The AI Market

There is a quiet reckoning happening in the AI industry. After years of launching every possible feature and chasing every possible use case, the biggest players are starting to pull back. The lesson is simple: if you scatter a product across ten different apps, one of two things happens. Users either get confused and stop using it, or they grab the fragment that works and ignore the rest. That is exactly what is forcing a round of consolidation right now.

OpenAI Simplifies, Microsoft Unifies

OpenAI’s leadership has acknowledged that fragmentation has been a real cost. Fidji Simo, OpenAI’s CEO of Applications, told employees that new bets like Codex forced the company to refocus. The net result is a desktop superapp that merges ChatGPT, the Codex assistant, and the Atlas browser. Meanwhile, Microsoft is building a similar one-stop shop for its Copilot products. The new app would connect GitHub Copilot, the consumer Copilot chatbot, Copilot Cowork, and an internal agentic-workflow project codenamed Autopilot. The idea is to give customers one coherent Copilot identity instead of letting personal and enterprise versions live in separate worlds.

That is a meaningful shift. Until recently, Microsoft had split teams into distinct consumer and commercial organizations, which made it hard to present a single vision. The new head of Copilot, Jacob Andreou, was hired from Snap with the explicit mandate of uniting those two worlds. The stakes are high because Microsoft was an early and massive AI investor, landing a $13 billion partnership with OpenAI, and yet the Copilot brand still trails both Google and OpenAI in active users. Less than 4.5 percent of the 450 million Microsoft 365 customers pay for Copilot features. GitHub Copilot, with more than 4.7 million paying subscribers, is healthier, but it faces competition from Cursor and Anthropic’s Claude Code.

Computer Use Becomes Real

A separate but related development is the spread of agentic capabilities. OpenAI’s Codex launched a computer-use feature this week that lets the app see a user’s screen and perform tasks directly. Windows users are now getting access in addition to Mac users. Users can also manage and review Codex jobs while away from the computer using the ChatGPT app. That is a big deal because it turns an AI assistant into an AI operator. It is the difference between a chatbot that answers questions and a system that executes multi-step workflows on your behalf.

The trade-offs are real. Amazon admitted as much after an internal AI leaderboard backfired. Workers began assigning autonomous bots to meaningless tasks just to climb a usage ranking, prompting an Amazon executive to tell staff not to use AI for its own sake. That is a useful reminder that capability without purpose creates noise, not value.

Regulation Catches Up, Fragile Models Get Cut

On the policy side, Illinois is about to sign an AI safety bill whose requirements go beyond anything New York or California has passed. The new rules demand independent audits and whistleblower protections at AI companies. Illinois Governor JB Pritzker has made clear he intends to sign the measure. It reflects a broader pattern: states are filling the federal void with layered, competing requirements. For any company shipping AI across state lines, compliance is becoming its own specialty.

Inside product labs, older model architectures are also being retired. OpenAI is sunsetting the Canvas interface for newer model generations, citing readability improvements in GPT-5.5 Instant. Canvas had allowed side-by-side editing of code and text, but users are being directed toward core chat experiences as the default. It is a small but telling example of how companies are prioritizing focus over breadth.

The EV Market

If AI vendors are consolidating, the EV industry is doing the opposite: it is pulling back. The past week included several decisions that exposed how uneven the transition to electric vehicles still is.

Tesla’s FSD Problem Is Bigger Than Advertised

A Reuters investigation into Tesla’s Full Self-Driving program found that many of the company’s own AI trainers do not trust the technology they are labeling. Employees reviewed footage of Teslas exceeding speed limits by twenty to thirty miles per hour after the introduction of an aggressive “Mad Max” driving mode. One labeler reported seeing an FSD-piloted vehicle traveling sixty mph in a twenty-five mph zone. Inside the company, workers routinely review video clips of animal deaths and narrow misses with children.

Tesla has long argued that its statistical methodology makes FSD safe by the numbers, but the investigation suggests there is a gap between what the metrics show and what the video evidence reveals. That gap matters because regulatory agencies, investors, and customers all rely on those claims when deciding how much autonomy the market can safely absorb.

Lexus Pauses, Lamborghini Chooses Hybrids, Rivian Pushes R2

In corporate strategy news, Toyota is discontinuing development of the mass-production version of its LF-ZC Lexus EV, the model originally meant to launch in 2026. Instead, Toyota is redirecting engineering resources toward SUVs. The move echoes last year’s decision by Lamborghini to cancel its EV plans in favor of plug-in hybrids. Lamborghini CEO Stephan Winkelmann told CNBC that hybrid strategy was the right move because customer acceptance curves for full EVs were not rising as expected. Ferrari also unveiled the Luce EV this week, but the car drew immediate backlash over its styling, especially given Jony Ive’s involvement in its design.

Not every EV story this week was a retreat. Waymo announced that its self-driving cars have entered Virginia and are mapping Arlington and Alexandria, even though the state does not yet permit autonomous operation. The company is preparing for future legislation while expanding its geographic footprint toward Washington, D.C. Rivian, meanwhile, is opening R2 order invitations on June 9th, starting with a Performance Launch Package model. Current R1T and R1S owners get priority access.

What the Retreats Mean

The pattern is not a rejection of electric vehicles so much as a correction of unrealistic timelines. Legacy automakers are being forced to admit that consumer demand, charging infrastructure, and battery economics do not line up as smoothly as five-year plans once assumed. The companies that survive this correction will be the ones with diversified lineups and realistic product cycles.

Companies, Products, and People Worth Watching

Anthropic and Claude Code

Anthropic’s Claude Code has emerged as one of the fastest-growing developer tools in the AI space. It is razor-focused on code generation and is winning converts who once depended on traditional IDEs. The company’s simplicity is a feature; by doing fewer things well, Claude Code avoids the sprawl that is currently forcing OpenAI and Microsoft to clean up their products. For developers who want an AI pair programmer without the confusion of a superapp, Claude Code is the clearest alternative.

Google’s Gemini and Workspace Integration

Google is deepening Gemini into Workspace for the enterprise market. A new feature rolling out on June 3rd lets users share Gemini conversations and canvases through Google Drive. Shared chats remain editable by collaborators without altering the original owner’s thread. It is a smaller move than the Microsoft or OpenAI superapp announcements, but it shows Google is treating Gemini as a first-class Workspace feature rather than an experimental sidebar.

Copilot Health AI

Microsoft also launched a preview of Copilot Health AI that can analyze personal medical records. The preview consolidates health data into a Copilot-driven interface. It is a reminder that the company’s AI ambitions extend well beyond office productivity and coding into personal healthcare. Expect more health-specific models from Microsoft, Google, and Apple over the next eighteen months.

The Energy Story the AI World Is Underwriting

None of the AI build-out is happening in clean-power vacuum. A set of recent IPOs tells the story. Fervo Energy, the enhanced geothermal company, went public in mid-May and now carries a market cap around $12.4 billion. X-energy, a small modular nuclear reactor developer, followed in April with an $11.5 billion market cap. Solv Energy, a utility-scale solar and battery company, raised a $6 billion valuation in February. All three are moving aggressively because AI data centers are pushing electricity demand up sharply after years of relative flatness.

The numbers are striking. Fervo’s Cape Station project in Utah targets five hundred megawatts, with first power expected before the end of October and additional units by early 2027. The company has binding power purchase agreements for more than six hundred megawatts and leases covering land capable of generating forty gigawatts. That is ten times the entire existing geothermal fleet in the United States.

Who Is Actually Building

Solv Energy already has twenty-one gigawatts operational across thirty-five states. Its solar-plus-storage model is one of the fastest ways to add capacity to the grid, which is why data-center operators keep showing up in its filings. X-energy is further from commercial reality, but its high-temperature gas-cooled reactors are designed to deliver eighty megawatts each with passive safety characteristics that regulators find easier to evaluate. The company received an important environmental approval last month for its Texas plant and is waiting on final Nuclear Regulatory Commission clearance.

These companies are not charity cases. They are responding to a market signal generated by the same AI boom that is producing Copilot superapps and self-driving cars. The data-center build-out is an infrastructure story as much as it is a software story.

Why It All Matters Right Now

The common thread across this week’s tech news is prioritization. AI companies are choosing consolidation over expansion. Automakers are choosing realism over promise. Energy companies are choosing scale because they know the AI industry is counting on it. In every case, the market is asking a simple question: what can you actually deliver, and at what cost?

The answer determines who gets funded, who gets regulated, and who gets left behind. The AI bargains being cut today are not about chatbots. They are about which products survive the transition from experiment to infrastructure. That is the story worth following.