Latest AI News
Marked-up Mac minis flood eBay amid shortages driven by AI
Overpriced Mac minis areflooding eBayamid shortages of the sold-out machines, which have become a favored tool for running on-device AI models likeOpenClaw. This week,reportsindicated that the $599 M4 Mac mini base model with 16GB RAM and 256GB of storage is sold out on Apple’s retail website, with no options for delivery or in-store pickup. The shortages have sinceextended to other configurationsof the base model, regardless of the amount of memory selected. This is the first time the base model has been sold out, someoutletsnoted. Meanwhile, models with higher storage (512GB and up) are only available to ship starting in June. As a result, eBay has become a secondary market for these in-demand computers. On the site, various configurations of the M4 Mac mini are available for sale at higher prices than if buying direct from Apple, which is no longer an option. Apple’s power-efficient Mac minis have become popular devices for testing and running at-home, on-device AI models, beginning with the OpenClaw craze but now extending to OpenClaw alternatives like ZeroClaw, other AI tools from Anthropic and OpenAI,Perplexity Computer,or otherspecialized local models.Unlike some PCs, Mac minis also run quietly and tend to be more reliable for 24/7 use, compared with laptop computers. The shortage of the devices also comes alongside anindustry-wide memory crunchand plans for a Mac mini refresh, according toBloomberg. However, refreshes of product lines haven’t led to shortages before. Apple did not immediately respond to a request for comment. This perfect storm of supply chain stress and increased demand for AI-friendly machines has inflated the prices of used consumer electronics. As of Friday morning, M4 base models with the 16GB RAM/256GB SSD configuration were selling at markups like $715-$795 for a new, “open box” model, and as high as $979 for an “excellent” refurbished version. Some “lightly used, pre-owned” Mac minis with this configuration were selling for around $700 — more than $100 more than the price of a new base model. There was also a single listing for a $925 brand-new M4 Mac mini with the same 16GB RAM and 256GB storage; the listing warned in bright red text: “Last one.” While you still may be able to score a reasonably priced refurb if you keep a close eye out (or if you win an eBay auction where the bid has started at a lower price point), it seems that the demand for the device is going to keep prices up until Apple’s supply chain refreshes. And now that the Mac mini is unavailable, Apple has begun to see increased demand for the Mac Studio, too. That computer isalso now sold outacross several configurations. AsArs Technicapointed out, you can still get a MacBook Pro with 128GB RAM and larger SSDs within a few weeks, and even the new and popular MacBook Neo is still shipping within two to three weeks. This suggests the real issue is consumer demand for the Mac mini itself.
View
Apple’s new CEO, and why Elon Musk wants to buy Cursor for $60B
A new era is on the way for Apple as Tim Cook plans to step down from his CEO role in September,handing the reins to hardware chief John Ternus. Ternus may be inheriting one of the most durable businesses in tech, but he’s also stepping into a very different ecosystem than the one Cook spent decades shaping. The App Store’s 30% cut is under pressure, the behind-the-scenes power Apple once held over developers is being challenged, andvibe-coded apps are changing what it means to build on Apple’s platform. On this episode of TechCrunch’sEquitypodcast, hosts Kirsten Korosec, Anthony Ha, and Sean O’Kane dig into what this transition means for startups and a closer look at some of the week’s biggest deals — including SpaceX’s $60B option on Cursor. Listen to the full episode to hear about: Subscribe to Equity onYouTube,Apple Podcasts,Overcast,Spotifyand all the casts. You also can follow Equity onXandThreads, at @EquityPod.
View
Google to invest up to $40B in Anthropic in cash and compute
Google plans to invest up to $40 billion in Anthropic and support the AI firm’s growing computing needs,Bloomberg reports. The Alphabet subsidiary is committing to invest $10 billion now, at a $350 billion valuation for Anthropic, with another $30 billion to follow if Anthropic hits certain performance targets, according to Anthropic. The promise of investment comes after Anthropic released its latest model, Mythos, to a limited group of partners this month. Anthropic says that Mythos is the company’s most powerful model to date and has significant cybersecurity applications. Due to potential misuse, Anthropic has restricted broader access while it works with select organizations to evaluate and address those risks — though the model has already fallen intounsanctioned hands. It’s also likely expensive to run at scale. The AI race is increasingly defined by access to the compute needed to train and deploy these systems. OpenAI has moved aggressively to secure that capacity through a web of multi-hundred-billion-dollar deals across cloud providers, chip suppliers, and energy, including anexpanded deal with chipmaker Cerebras this month. Anthropic has been in a scramble of its own. The company has facedwidespread complaintsabout Claude use limits in recent weeks and responded with a bevy of infrastructure deals. Earlier this month, Anthropicstruck a dealwith cloud computing provider CoreWeave for data center capacity. It also this week secured anadditional $5 billion investmentfrom Amazon, part of a broad agreement under which Anthropic is expected to spend up to $100 billion for around 5 gigawatts of compute capacity over time. While Google is a direct competitor in AI models, it’s also a key infrastructure supplier to Anthropic. Anthropic relies heavily on Google Cloud for chips and infrastructure, including access to Google’s tensor processing units (or TPUs), which are specialized chips designed for AI workloads and considered among the best alternatives to Nvidia’s in-demand processors. Anthropic’s relationship with Google predates this week’s news.Earlier this month, Anthropic announced a partnership with Google and chipmaker Broadcom, which designs custom AI chips for Google, to access multiple gigawatts of TPU-based computing capacity beginning in 2027; a subsequent Broadcom securities filing put that figure at 3.5 gigawatts. The new Google investment expands that arrangement, with Google Cloud now providing a fresh 5 gigawatts of capacity over the next five years, with room to scale further. Anthropic’s valuation stood at $350 billion as recently as February; investors have since been eager to back the company at $800 billion or more, according to Bloomberg. The company is also reportedly considering an IPO as soon as October.
View
ComfyUI hits $500M valuation as creators seek more control over AI-generated media
ComfyUI, a startup that helps creators control image, video, and audio outputs from diffusion models with a node-based workflow, has raised a $30 million funding round at a $500 million valuation. The round was led by Craft Ventures, with participation from other investors including Pace Capital, Chemistry, and TruArrow. ComfyUI was started as an open-source project in 2023, shortly after the introduction of diffusion models. At that time, models like Midjourney and OpenAI’s DALL-E were barely functional, frequently making major mistakes, such as adding extra fingers to hands. To address these limitations, the project founders developed a modular framework that gives creators granular control over every step of the generation process. Their tool gained such significant traction among creative professionals that it eventually evolved into a formal startup. In late 2024, ComfyUI raised $19 million in Series A financing from investors including Chemistry Ventures, Cursor Capital, and Guillermo Rauch, founder of Vercel. Although the latest diffusion models have come a long way from adding a sixth digit to hands, the need for the granular precision that ComfyUI offers has only grown. “If you think about your typical prompt-based solution, like Midjourney or ChatGPT, you ask for something, it [gets only] 60% – 80% there,” Yoland Yan, ComfyUI’s co-founder and CEO, told TechCrunch. “But to change that remaining 20%, you have to try this slot machine.” Yan (pictured left) compared the process to playing in a casino because prompting the model to make a small change can result in a completely different output, including overwriting the parts that were already perfect. ComfyUI’s node-based interface allows creators to link specific components of the generation process, giving them full control over the quality of their final output. “You cannot easily convey that message in the prompt box [of a foundational model],” Yan said. Creators seem to agree, as ComfyUI claims to have over 4 million users. The tool is being used by creative professionals for visual effects, animation, advertising, and even industrial design. The startup says its offering has become such a necessary tool of the trade for technical artists and other creatives that it is not uncommon to see “ComfyUI artist or engineer” listed as a job title on studio job boards. Although video and image foundational models continue to improve, Yan claims that they are far from perfect, and a tool like ComfyUI will continue to be in high demand. “In the world where AI slop is going to be everywhere, the Comfy version of human-in-the -loop approach is going to win out most of the eyeballs in the end,” he said. ComfyUI’s competitors includeWeavy, a startup that was acquired by Figma last year.
View
DeepSeek previews new AI model that ‘closes the gap’ with frontier models
Chinese AI lab DeepSeek has launched two preview versions of its newest large language model,DeepSeek V4, a much-awaited update to last year’s V3.2 model and the accompanyingR1 reasoning modelthat took theAI world by storm. The company says both DeepSeek V4 Flash and V4 Pro are mixture-of-experts models with context windows of 1 million tokens each — enough to allow large codebases or documents to be used in prompts. The mixture-of-experts approach involves activating only a certain number of parameters per task to lower inference costs. The Pro model has a total of 1.6 trillion parameters (49 billion active), which makes it the biggest open-weight model available, outstripping Moonshot AI’s Kimi K 2.6 (1.1 trillion), MiniMax’s M1 (456 billion), and more than double DeepSeek V3.2 (671 billion). The smaller, V4 Flash has 284 billion parameters (13 billion active). DeepSeek says both models are more efficient and performant than DeepSeek V3.2 due to architectural improvements, and have almost “closed the gap” with current leading models, both open and closed, on reasoning benchmarks. The company claims its new V4-Pro-Max model outperforms its opensource peers across reasoning benchmarks, and outstrips OpenAI’s GPT-5.2 and Gemini 3.0 Pro on some tasks. In coding competition benchmarks, DeepSeek said both V4 models’ performance is “comparable to GPT-5.4.” However, the models seem to fall slightly behind frontier models in knowledge tests, specifically OpenAI’s GPT-5.4 and Google’s latest Gemini 3.1 Pro. This lag suggests a “developmental trajectory that trails state-of-the-art frontier models by approximately 3 to 6 months,” the lab wrote. Both V4 Flash and V4 Pro support text only, unlike many of its closed-source peers, which offer support for understanding and generating audio, video, and images. Notably, DeepSeek V4 is much more affordable than any frontier model available today. The smaller V4 Flash model costs $0.14 per million input tokens and $0.28 per million output tokens, undercutting GPT-5.4 Nano, Gemini 3.1 Flash, GPT-5.4 Mini, and Claude Haiku 4.5. The larger V4 Pro model, meanwhile, costs $0.145 per million input tokens and $3.48 per million output tokens, also undercutting Gemini 3.1 Pro, GPT-5.5, Claude Opus 4.7, and GPT-5.4. The launch comes a day after the U.S.accusedChina of stealing American AI labs’ IP on an industrial scale using thousands of proxy accounts. DeepSeek itself has been accused by Anthropic and OpenAI of “distilling,” essentially copying, their AI models.
View
Nothing introduces an AI-powered dictation tool
In the last few years, AI-powered dictation tools have taken off. In addition to existing dictation apps like Wispr Flow, Superwhisper, Willow, and Monologue,newonesare being launched every week. On Thursday, hardware company Nothing launched a competitive product of its own, called Essential Voice. The core idea is similar to other dictation apps, as Essential Voice works in any app to turn your speech into formatted text, removing filler words like “um” and “ah” along the way. The company said that you can also create custom voice shortcuts for words, links, templates, and repeated phrases. For instance, you can assign the “my address” shortcut to your full address. At the moment, the feature is available on the Phone (3) with rollout for Phone (4a) Pro planned for later this month, and support for Phone (4a) arriving next month. The average person types 36 words a minute on a phone.But, they can say it four times faster.Essential Voice turns your speech into clear, ready-to-use writing.pic.twitter.com/l08bnS8sNF To access the feature, users either press the Essential key on devices where it is present or activate it from the keyboard. The feature issimilar to the one that Superwhisperreleased earlier this week for iPhone users, which allows them to map the iPhone’s action key to the app’s keyboard for dictation. Nothing’s new tool can also translate text directly from one language to another. At launch, Nothing said the feature supports over 100 languages. Going forward, it will also introduce app-based custom styling, meaning you’ll be able to change the tone of the AI editing within app categories, like work and messaging. Nothing is one of the first companies to offer a system-level integration for dictation. However, based onGoogle’s recent release of its offline dictation app, we might see more companies release similar tools in the future.
View
Uber CTO Praveen Neppalli Naga joins stacked StrictlyVC SF lineup for April 30 event
Surprise!StrictlyVC San Francisco, which will kick off this year’s events lineup for TechCrunch on April 30 at the Sentro Filipino Cultural Center, is getting a new addition to its increasingly stacked roster of speakers. Uber CTO Praveen Neppalli Naga will join us to discuss, you guessed it, operating at scale in the age of AI. You’ll need to act swiftly to grab a ticket, though, for what’s becoming the go-to event within the SF startup scene next week. It’s an ideal one-stop shop for any founder or investor looking to widen their networks, deepen community ties, and learn from what’s now a wildly deep roster of speakers. Naga’s conversation with TechCrunch editor-in-chief Connie Loizos will cover that wide purview, exploring what it’s taken to build the many complex, interwoven systems amid the AI revolution on one of the most widely used services on the planet. And his work with Uber runs deep, having been with the company since 2015, long preceding the AI boom that is refining what CTOs have to focus on. And it’s notallAI. Naga has had a particular focus on developing earnings systems for drivers and couriers within Uber’s network, and he had previously played a key role in building LinkedIn’s early products and infrastructure that set it up to be the mainstay of professional life it is today. For those who haven’t been keeping score, we now have five (!!) speakers, including Naga: Eclipse founder and CEOLior Susan, whose recently raised $1.3 billion fund is focusing on physical AI startups, will take a deep dive into the kinds of ventures and projects that excite him and that other investors should take note of. Replit co-founder and CEOAmjad Masadwill provide a look into the future of AI-driven software development, which couldn’t be coming at a better time amid big claims about AI coding capabilities and major concerns from engineers. TDK Ventures, our sponsor for the evening, will have its president,Nicolas Sauvage, host an essential conversation about the ins and outs of raising capital from strategic backers and other early-stage investing topics that founders and VCs won’t want to miss. AndCampbell Brown, former CNN host and Meta media partnerships lead, will share her stories from entering the startup community with Forum AI, as she looks to contribute to the efforts to stem disinformation and inaccurate claims that arise from the misuse of AI. Don’t wait, don’t procrastinate,act now and snag a ticketbefore word on our latest speakers gets around. Block your calendar and make time to join the StrictlyVC community, and set yourself up for future success with the lessons learned from this SF event!
View
India on the move: Personalisation, Sustainability, AI, and the New Era of Hospitality
Overall, 89% of Indian travellers say they want personalised trips, compared with 74% globally.
View
Data Centre in a Box: The Future of AI Infra May Be Prefabricated
As AI drives an unprecedented surge in computing demand, modular and prefabricated data centres are emerging as efficient solutions.
View
Microsoft Makes Copilot’s Agentic Features in Word, Excel and PowerPoint Generally Available
Microsoft made Copilot's agentic capabilities generally available in Word, Excel, and PowerPoint on Wednesday. The new features are being expanded to the different paid tiers of Microsoft 365, and the agentic features will be set as the default. The Redmond-based tech giant said these features will allow Copilot to not only answer queries, but also take action directly on the document. While Copilot will be the main chatbot, users will be able to pick from a large selection of third-party artificial intelligence (AI) models.
View
Adobe Previews New Agentic AI Workflows for Marketing Tasks at Adobe Summit 2026
Adobe shared several previews of how its artificial intelligence (AI) and agentic features can help transform marketing workflows. These previews were showcased at the Adobe Summit 2026 during the Sneaks sessions, where the company presents early-stage projects that are currently in the research and development (R&D) stage. These are built by the software giant's scientists, engineers, product managers, and designers. This year, the focus was on marketing workflows and how AI can reimagine content production and branded experiences.
View
In another wild turn for AI chips, Meta signs deal for millions of Amazon AI CPUs
Amazon just scored a major coup with Meta thanks, once again, to Amazon’s own homegrown chips. Meta has signed a deal to use millions ofAWS Gravitonchips to power its growing AI needs, Amazon announced Friday. Note that the AWS Graviton is an ARM-based CPU, (a central processing unit, the chip that handles general computing tasks) not a GPU (a graphical processing unit). While GPUs remain the chip of choice for training large models, once those models are trained, AI agents built on top of them are causing a shift in the type of chip is needed. Agents create compute-intensive workloads like real-time reasoning, writing code, search, and the the coordination involved in managing agents through multi-step tasks. AWS’s latest version of Graviton was designed specifically to handle AI-related compute needs, the company says. This deal brings more of Meta’s cash back to AWS instead of competitors like Google Cloud. Last August,Meta signed a six year, $10 billion deal with Google Cloud, though Meta had, until then, primarily been an AWS customer that also used Microsoft Azure. We couldn’t help but notice that AWS timed the announcement of this deal right as the Google Cloud Next conference wrapped up, like a virtual smirk at its cloud rival. Google, of course, also makes its own custom AI chips andannounced new versions of them at the show. True, Amazon makes its own AI GPU as well: the Trainium, which, despite its name, is used for both training and inference — the stage that happens after a model is trained, when it’s actively processing prompts. But Anthropic had already swooped in with a deal announced earlierthis month that commandeered many of those chipsfor years to come. The Claude maker agreed to spend $100 billion over 10 years to run its workloads on AWS — with a particular focus on Trainium — while Amazon agreed to invest another $5 billion (bringing its total to $13 billion of investment) into Anthropic in return. Ultimately, the Meta deal is allowing Amazon to showcase a huge AI customer as a proving point for its homegrown CPUs. These are chips that compete with Nvidia’s new Vera CPU, which is also ARM-based and designed to handle AI agentic workloads. The difference, of course, is that Nvidia sells its chips and AI systems to enterprises and cloud providers (including AWS). AWS only sells access to its chips through its cloud service. Earlier this month Amazon CEO Andy Jassytook aim at Nvidia and Intel in his annual shareholder letter, saying that enterprises want better price-performance ratios for AI, and that he intends to win deals on that basis. This also means the pressure couldn’t be higher on Amazon’s internal chip building team to deliver, a team thatwe visited last month in an exclusive tour of their lab.
View