★ Top Story
TechCrunch AINvidia reported yet another record revenue quarter and disclosed $43 billion in startup holdings, signaling the depth of its strategic bets across the AI ecosystem. CEO Jensen Huang separately identified a "brand new" $200 billion opportunity in CPUs purpose-built for AI agents, framing it as Nvidia's next major growth frontier. Revenue growth is expected to moderate next quarter, but Huang's agent-CPU thesis suggests the company is already positioning for the next hardware wave.
Read the original →
TechCrunch AIAnthropic has informed investors it expects to more than double revenue to approximately $10.9 billion in its second quarter, marking what would be the company's first profitable period. The milestone is a significant inflection point for one of the most heavily funded AI labs, which has burned through billions in compute costs since its founding. A newly disclosed $1.25 billion-per-month compute deal with xAI underscores both the scale of Anthropic's infrastructure needs and the unusual cross-competitor dynamics now shaping the AI industry.
Read more →
TechCrunch AIDetails buried in SpaceX's IPO filing reveal that Anthropic has agreed to pay Elon Musk's xAI $1.25 billion per month for compute capacity — a stunning arrangement between two ostensibly competing AI labs. The deal highlights the severe shortage of high-quality GPU infrastructure and the lengths frontier AI companies will go to secure it. It also provides xAI with a substantial and predictable revenue stream as it continues to scale its Grok models.
Read more →
TechCrunch AISpaceX's IPO prospectus has provided the first public window into xAI's finances, revealing the company lost $6.4 billion in 2025 while aggressively expanding its Grok AI platform. The filing also discloses that xAI is purchasing $2.8 billion worth of natural gas turbines over three years to power its data centers — even as the company faces lawsuits over environmental concerns related to its existing generators. The disclosures paint a picture of a company spending at extraordinary velocity with no near-term path to profitability.
Read more →
The VergeSpaceX's long-awaited IPO filing has landed, and it does more than open the books on the rocket company — it exposes the tangled financial web connecting Musk's ventures including Tesla, xAI, and SpaceX itself. Musk is explicitly named as a risk factor, with the filing detailing how money and resources flow between his companies in ways that could create conflicts of interest for public shareholders. The offering could make Musk the world's first trillionaire if valuations hold.
Read more →
TechCrunch AIOpenAI says its latest reasoning model has disproved a geometry conjecture that has stood unsolved since 1946 — and this time, the independent mathematicians who previously called out the company for a false claim are backing the result. The announcement represents a meaningful benchmark for AI's capacity to contribute genuine scientific discovery rather than merely pattern-match on known proofs. If verified broadly, it would mark one of the most credible demonstrations yet of AI-driven mathematical research.
Read more →
TechCrunch AIOne day after a jury swiftly rejected Elon Musk's lawsuit challenging OpenAI's corporate structure, the company is reportedly accelerating preparations for a public offering potentially as early as September. The lawsuit had cast uncertainty over OpenAI's ongoing conversion from a nonprofit to a for-profit entity, and its dismissal clears a significant legal obstacle. The IPO would be one of the most consequential tech listings in years, given OpenAI's central role in the current AI boom.
Read more →
TechCrunch AIThe jury's rapid verdict against Elon Musk in his lawsuit against OpenAI, Sam Altman, and Microsoft reflected what courtroom observers described as a fundamentally weak case. Evidence presented at trial suggested Musk himself had harbored ambitions to control OpenAI's direction and resources, undermining his central claim that Altman had betrayed the company's nonprofit mission. MIT Technology Review's detailed trial coverage reinforces that the outcome was less a vindication of OpenAI's conduct than a rejection of Musk's standing to challenge it.
Read more →
Google AI BlogGoogle used I/O 2026 to unveil Gemini 3.5, a new frontier model series combining reasoning with real-world action, alongside a sweeping overhaul of Search powered by agentic AI that proactively monitors topics and surfaces updates for users. The company also announced Gemini Omni integration across YouTube Shorts, Gmail voice search, and a new design tool called Google Pics, signaling a full-stack push to embed AI into every consumer surface. DeepMind CEO Demis Hassabis closed the keynote by suggesting the industry may be entering the "foothills of the singularity" — a claim that drew immediate scrutiny and debate.
Read more →
Ars TechnicaGoogle's AI Mode, now one year old, is shifting user behavior from keyword queries to natural language conversations, and the company is doubling down with agentic "information agents" that run in the background and proactively alert users to relevant changes. The evolution of Search also comes with an expanded advertising layer: Gemini AI now generates custom product explainers alongside shopping results, embedding sponsored content more deeply into AI-generated responses. Critics and analysts are watching closely to see whether the ad integration undermines user trust in AI-driven answers.
Read more →
Ars TechnicaFigure AI's around-the-clock livestream of its humanoid robots autonomously sorting and handling packages has gone viral, drawing millions of viewers fascinated by the fluid, human-like movements. The footage has reignited public conversation about the pace of physical AI development and how close warehouse automation is to mainstream deployment. The viral moment also underscores a broader cultural phenomenon: humans appear to have a deep, instinctive engagement with robots that look and move like people.
Read more →
The VergeCounty commissioners in Box Elder County, Utah have approved the Stratos Project, a proposed 40,000-acre data center complex in Hansel Valley billed as a cornerstone of American AI infrastructure dominance. The project has drawn sharp opposition from local residents, environmental experts, and water-use advocates who warn of severe strain on the region's scarce resources. The approval highlights the growing tension between the AI industry's insatiable appetite for compute capacity and the communities bearing the environmental and social costs.
Read more →
The VergeGoogle's SynthID and the C2PA Content Credentials standard are undergoing their largest coordinated expansion to date, embedding invisible provenance metadata into AI-generated images, video, and audio at scale. The rollout is being watched as a critical test of whether technical watermarking solutions can meaningfully combat deepfakes and synthetic media disinformation in real-world conditions. Skeptics note that determined bad actors can strip or spoof metadata, and that adoption gaps across platforms remain a fundamental weakness.
Read more →
Ars TechnicaChinese customs authorities added Nvidia's RTX 5090D V2 to a list of banned imports last Friday — the same week CEO Jensen Huang was in the country on a high-profile visit. The move signals that Beijing is continuing to tighten restrictions on advanced Nvidia hardware despite ongoing diplomatic and commercial engagement, and it puts Nvidia in the awkward position of courting a market that is simultaneously closing its doors to the company's most capable chips. The ban adds fresh complexity to Nvidia's China strategy at a moment when the company is navigating U.S. export controls from the other direction as well.
Read more →
TechCrunch AIFigma has launched a native AI agent within its collaborative design environment, allowing users to generate new designs, edit existing ones, and automate repetitive tasks using natural language prompts. The move positions Figma directly against Google's new AI design tools announced at I/O 2026, intensifying competition in what is rapidly becoming one of the most contested segments of the creative software market. The integration reflects a broader industry shift toward AI agents embedded inside professional tools rather than offered as standalone products.
Read more →
TechCrunch AIStability AI has launched Stability Audio 3.0, a new generative audio model capable of producing tracks up to six minutes long, with a smaller variant designed to run entirely on-device. The release marks a notable step forward in audio generation quality and accessibility, particularly for creators who need low-latency, privacy-preserving music generation without cloud dependency. It also signals that Stability AI, despite its turbulent recent history, remains an active competitor in the generative media space.
Read more →No stories from this source today.