Nvidia AI Chip Price Hike & Anthropic Builds Own Silicon

Samantha: Welcome to The State of Tech, The European Edition, Sunday August twenty-three, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Nvidia hikes AI chip prices by more than fifteen percent, STMicroelectronics pushes through its third price rise of the year, Anthropic hires a Google chip veteran to build its own silicon, an African startup brings voice AI to seventy-five languages by phone call, and to close out, two things you can actually use today, the free open-source screen recorder ScreenToGif and the offline reading app Pocketbook Reader. Let's get into it.

Nvidia set to raise AI chip prices by more than fifteen percent, squeezing European cloud budgets.

Samantha: We start with the cost of the AI boom, and it is going up again. Nvidia is telling server makers that its next wave of AI systems, the ones built around its Vera Rubin and Grace Blackwell chips, will be more than fifteen percent more expensive from early 2027. The reason, according to the reporting, is not greed but memory. The high-bandwidth memory that sits next to these chips has become scarce and pricey, and Nvidia is passing that cost straight through to its customers.

Bob: And those customers are the big server builders that supply Microsoft, Google and Oracle. They have already started warning hyperscale clients that the bill for the next generation of AI infrastructure is going to be noticeably heavier. Fifteen percent on a single accelerator sounds manageable, but when you are ordering tens of thousands of them for a new data centre, it adds up to hundreds of millions of extra spending per site. And that is before power, cooling and real estate.

Samantha: The bottom line for Europe is that this lands right in the middle of the sovereign cloud debate. European providers like OVHcloud, Deutsche Telekom's cloud arm and Scaleway are trying to build credible AI offerings without leaning entirely on the American hyperscalers. Higher chip prices make that harder, because smaller buyers get worse deals than Microsoft or Amazon do. So the gap between European alternatives and the giants risks widening, not shrinking.

Bob: There is also a knock-on effect for anyone renting AI capacity. Startups training their own models, universities running research clusters, banks fine-tuning private assistants, they all pay per hour of GPU time. If the underlying hardware is fifteen percent more expensive, those rates do not stay flat for long. Expect European AI startups to feel this in their 2027 budgets, and expect more of them to push harder on efficiency, smaller models, and shared infrastructure to keep costs under control.

STMicroelectronics pushes through its third price hike of the year for semiconductors.

Samantha: Staying with the theme of rising costs, but well outside the AI bubble. STMicroelectronics, the Franco-Italian chipmaker, put its third price increase of 2026 into effect today. Various product lines, from power chips to microcontrollers, are going up. The company points to strong demand, but also to higher costs for transport, energy, raw materials and manufacturing services. And this is not isolated: NXP, onsemi and Infineon have all moved in the same direction this year.

Bob: What makes this story interesting is where these chips actually end up. STMicroelectronics parts sit inside cars, industrial machines, home appliances, medical devices and factory automation. These are not the glamorous AI accelerators, they are the workhorses of European industry. When their prices go up three times in a year, that pressure filters through to carmakers in Germany, appliance builders in Italy and industrial equipment makers across the continent.

Samantha: The bigger picture is that the chip industry is running two speeds. The AI end is red hot, with buyers willing to pay almost anything for advanced accelerators. The mature end, the chips that keep a washing machine or a hybrid car running, is supposed to be the calmer market. But energy costs in Europe, geopolitical friction, and stretched supply chains mean even that segment is not calm anymore. For European manufacturers, this is a slow squeeze on margins that does not make headlines but shows up in every quarterly report.

Bob: And for consumers, it is a reminder that the price of a new car, a new dishwasher or a new industrial robot is being shaped by decisions made in chip factories, not showrooms. Industry groups have been arguing for years that Europe needs more of its own chip production. Days like this make that argument concrete.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Anthropic hires Google chip veteran to design its own custom AI silicon.

Bob: Our third story is about who controls the hardware behind the biggest AI models. Anthropic, the company behind the Claude assistant, has hired Amir Salek, one of the founders of Google's custom chip programme. Salek previously led Google's Tensor Processing Unit business, the in-house chips that Google has used for years to train and run its own AI. He is joining Anthropic's compute team, and the signal is unmistakable: Anthropic wants its own silicon.

Samantha: This lines up with a broader trend. OpenAI is reportedly working on custom chips with Broadcom. Meta and Amazon already design their own accelerators. Google has been at it the longest with TPUs. The reason is simple. Depending on Nvidia is expensive, and, as we just heard in the first story, getting more expensive. If you are training frontier models that cost hundreds of millions of dollars per run, even small efficiency gains from custom chips translate into real savings and real independence.

Bob: The catch is that designing chips is genuinely hard. It takes years, deep engineering talent, and access to advanced manufacturing at TSMC or Samsung. Hiring one senior leader does not produce silicon overnight. But it does set the direction. Anthropic is telling investors and partners that it plans to be a full-stack AI company, not just a model developer renting time on someone else's hardware.

Samantha: For Europe, the notable thing is what is missing from this list. The big custom AI chip efforts are all American. European chip design talent exists at companies like ARM in the UK, at Imec in Belgium, at research groups across France and Germany, but there is no equivalent European AI lab building its own accelerators at scale. That is a strategic gap, and one the European Chips Act was supposed to help close. Stories like this show how far there still is to go.

African startup EqualyzAI brings voice AI to seventy-five languages, over an ordinary phone call.

Bob: Our fourth story shifts the frame entirely. EqualyzAI, a startup founded in Africa, is building voice-first AI that understands more than seventy-five African languages. The clever bit is not just the language coverage, it is the delivery. Users interact through a regular phone call, no smartphone or data plan needed. In many African countries, mobile phone coverage vastly exceeds reliable internet access, so meeting people on a voice call is meeting them where they actually are.

Samantha: The company is targeting healthcare, agriculture, education and financial inclusion. Picture a farmer calling a number to ask about crop disease in their own language, or a patient describing symptoms and getting triage advice, or a small trader checking market prices, all through spoken conversation rather than an app. Most global AI development assumes English, a smartphone, and a fast connection. EqualyzAI is throwing those assumptions out and asking what AI looks like when none of them apply.

Bob: For European audiences, this matters in two ways. First, as a model. Europe has plenty of smaller languages and regional dialects that are poorly served by big American AI systems. Welsh, Basque, Sami, Frisian, and many others could benefit from the same voice-first, low-bandwidth approach. Second, as a reminder that AI inclusion is not just about smart glasses in Berlin, it is about who gets left out when the default is always the newest hardware and the fastest connection.

Samantha: There is also a development angle. European aid agencies and NGOs spend significant budgets on digital projects in African countries. Local-language, phone-based AI could deliver public health information, agricultural advice, or financial services at a fraction of the cost of building apps that most target users cannot access anyway. It is a quiet argument for supporting African AI companies rather than just exporting European or American tools.

ScreenToGif is the free open-source recorder that turns any screen into a shareable clip in seconds.

Bob: To close out, two things you can actually use today. First, ScreenToGif. It is a free open-source Windows tool that records a chunk of your screen, a webcam feed, or a sketchboard, and turns it into a GIF, an MP4, or an animated image you can paste into a chat, an email, or a bug report. If you have ever tried to explain to a colleague why some button is not working, or wanted to show a friend a funny moment in a video call, this is the tool that makes it painless.

Samantha: What makes it stand out is how lightweight it is. Users report it launches instantly, records only the window you point it at, and includes a proper editor afterwards. You can trim frames, add captions, blur out sensitive information, or draw arrows on top of the recording. All of that in an app that is genuinely free, with no watermark, no account, and no upload to a cloud service. Your recording stays on your own machine unless you decide to share it.

Bob: For European users, the privacy angle is real. A lot of screen recording tools either upload everything to a US-hosted service or push you into a subscription. ScreenToGif does neither. It is popular with developers filing bug reports, with teachers making short explainer clips, and with anyone tired of writing three paragraphs when a five second video would do the job. It runs on Windows 10 and 11, and installation is a quick download from the official site.

Samantha: One practical tip. If you are recording something for public sharing, use the built-in editor to strip out any browser tabs, notification pop-ups, or file names that reveal more than you meant to show. ScreenToGif makes that easy with its frame-by-frame editing. It is exactly the kind of tool that should be in everyone's back pocket, but somehow rarely is.

Pocketbook Reader is the free cross-platform e-book app that reads almost any format on any device.

Bob: Our second find is for readers. Pocketbook Reader is a free e-book app made by the European e-reader company Pocketbook. It runs on Android, iOS, Windows and macOS, and it handles just about every e-book format you can throw at it, EPUB, PDF, MOBI, FB2, and more. If you have a mixed collection of books from different shops and old downloads, this app opens them all in one clean interface.

Samantha: What makes it useful is that it does not lock you into one store. Unlike Amazon's Kindle app or Apple Books, Pocketbook Reader treats your library as your library. You can side-load books, sync between your phone and your laptop, and adjust fonts, spacing and background colours to whatever helps you read comfortably. It also supports audiobooks and text-to-speech, which is handy for long journeys.

Bob: For Europeans in particular, there is a nice bonus. Pocketbook is a European company, and the app integrates with its own e-book store which sells titles in dozens of European languages, including many smaller ones that are hard to find on the big American platforms. You do not have to use their store, but it is there if you want an alternative to Amazon.

Samantha: The app is free with no ads, and it works offline once your books are downloaded. Reviewers praise its clean interface and its willingness to open almost anything, including scanned PDFs that some competitors struggle with. If your reading is scattered across three different apps right now, Pocketbook Reader is a quiet way to pull it all back into one place. And that is our lot for today.

Bob: Time to wrap.

Samantha: Today we covered: Nvidia's fifteen percent AI chip price hike, STMicroelectronics' third price rise of the year, Anthropic hiring a Google chip veteran to build its own silicon, EqualyzAI bringing voice AI to seventy-five African languages by phone call, and two things to try yourself: ScreenToGif and Pocketbook Reader.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.