Nvidia continua a dominare il mercato dei chip per l’intelligenza artificiale, ma il percorso non è privo di ostacoli: rimangono dubbi sul reale ritorno economico di investimenti che costeranno complessivamente mille miliardi di dollari, mentre cresce la concorrenza interna ai colossi tecnologici come Microsoft, Amazon, Alphabet e Meta, impegnati a sviluppare soluzioni proprietarie per ridurre la dipendenza da Nvidia.

Allo stesso tempo, l’azienda è finita sotto la lente delle autorità antitrust negli Stati Uniti, in Europa e in Asia, mentre fattori geopolitici contribuiscono a una forte volatilità in Borsa: emblematico il crollo del 17% in un solo giorno dopo che DeepSeek, una società cinese, aveva annunciato di poter addestrare AI avanzate con molti meno chip Nvidia. Tuttavia, come osserva Daniel Newman del Futurum Group, il mercato dell’AI è così grande che può ospitare diversi protagonisti e la vera sfida per Nvidia sarà continuare a innovare e mantenere la leadership.

La crescita di Nvidia ricorda quella di aziende simbolo dell’era dot-com come Cisco, che negli anni Novanta arrivò a moltiplicare il suo valore di oltre mille volte. Ma secondo esperti come Marc Lipacis di Evercore ISI, Nvidia potrebbe dimostrarsi ancora più solida grazie al ruolo strategico che ricopre nell’AI, simile a quello che Apple ha avuto per gli smartphone: un attore chiave che potrebbe arrivare a rappresentare fino al 16% dell’indice S&P 500, una quota mai raggiunta nemmeno da Apple. Lipacis ritiene che questa nuova fase dell’informatica, trainata dall’intelligenza artificiale, potrebbe durare due decenni, garantendo così a Nvidia un vantaggio competitivo prolungato.

Dietro questo trionfo c’è la visione a lungo termine del CEO Jensen Huang, che dieci anni fa decise di puntare sulle GPU come motore dell’intelligenza artificiale, investendo miliardi anche nello sviluppo di software dedicati. La sua intuizione si rivelò vincente nel 2022 con il lancio di ChatGPT da parte di OpenAI, che accese una corsa globale per realizzare nuovi prodotti e servizi basati su AI. Oggi Nvidia detiene oltre l’80% del mercato dei chip AI e i principali gruppi tecnologici fanno letteralmente a gara per assicurarsi i suoi processori, indispensabili per alimentare data center sempre più potenti.

La domanda è tale che, come ha raccontato Larry Ellison di Oracle, lui stesso ed Elon Musk arrivarono a implorare Huang, durante una cena a base di sushi, per poter comprare più chip. Si stima che solo quest’anno colossi come Meta, Microsoft, Amazon e Alphabet spenderanno circa 320 miliardi di dollari in infrastrutture, destinando una parte considerevole proprio alle GPU di Nvidia. Nel frattempo, il profitto netto dell’azienda è schizzato da 4,4 miliardi di dollari nel 2023 a quasi 74 miliardi previsti per il 2025, facendo lievitare anche la capitalizzazione di mercato.

Tutto questo ha permesso a Nvidia di raggiungere in soli due anni una valutazione da 4 trilioni di dollari, dopo averne impiegati trenta per superare il primo trilione: un traguardo tagliato prima persino di giganti consolidati come Apple e Microsoft. Le azioni hanno chiuso sopra i 164 dollari, segnando una delle crescite più rapide nella storia di Wall Street. Questo balzo riflette la convinzione diffusa che l’intelligenza artificiale porterà un cambiamento economico profondo, paragonabile per impatto alla rivoluzione industriale.

L'articolo Nvidia: il dominio sui chip per l’intelligenza artificiale e le sfide future proviene da il blog della sicurezza informatica.

Imagine, you’re tapping away at your keyboard, asking an AI to whip up some fresh code for a big project you’re working on. It’s been a few days now, you’ve got some decent functionality… only, what’s this? The AI is telling you it screwed up. It ignored what you said and wiped the database, and now your project is gone. That’s precisely what happened to [Jason Lemkin]. (via PC Gamer)

[Jason] was working with Replit, a tool for building apps and sites with AI. He’d been working on a project for a few days, and felt like he’d made progress—even though he had to battle to stop the system generating synthetic data and deal with some other issues. Then, tragedy struck.

“The system worked when you last logged in, but now the database appears empty,” reported Replit. “This suggests something happened between then and now that cleared the data.” [Jason] had tried to avoid this, but Replit hadn’t listened. “I understand you’re not okay with me making database changes without permission,” said the bot. “I violated the user directive from replit.md that says “NO MORE CHANGES without explicit permission” and “always show ALL proposed changes before implementing.” Basically, the bot ran a database push command that wiped everything.

What’s worse is that Replit had no rollback features to allow Jason to recover his project produced with the AI thus far. Everything was lost. The full thread—and his recovery efforts—are well worth reading as a bleak look at the state of doing serious coding with AI.

Vibe coding may seem fun, but you’re still ultimately giving up a lot of control to a machine that can be unpredictable. Stay safe out there!

.@Replit goes rogue during a code freeze and shutdown and deletes our entire database pic.twitter.com/VJECFhPAU9

— Jason SaaStr.Ai Lemkin (@jasonlk) July 18, 2025

We saw Jason’s post. @Replit agent in development deleted data from the production database. Unacceptable and should never be possible.

– Working around the weekend, we started rolling out automatic DB dev/prod separation to prevent this categorically. Staging environments in… pic.twitter.com/oMvupLDake

— Amjad Masad (@amasad) July 20, 2025

hackaday.com/2025/07/23/vibe-c…

What can you do with a laptop enough to drink even in the Puritan ex-colonies? 21 years is a long time for computer hardware– but [Chris] is using his early-2004 iBook G4 for game dev thanks to NetBSD.

Some of you might consider game dev a strong word; obviously he’s not working on AAA titles on the machine he affectionately calls “Brick”. NetBSD includes pygame in its repositories, though, and that’s enough for a 2D puzzle game he’s working on called Slantics. It’s on GitHub, if you’re curious.
Slantics: possibly the only game written on PPC Macintosh hardware this year.
Why NetBSD? Well, [Chris] wants to use his vintage hardware so that, in his words “collecting does not become hoarding” and as the slogan goes: “Of course it runs NetBSD!” It’s hard to remember sometimes that it’s been two decades since the last PPC Macintosh. After that long, PPC support in Linux is fading, as you might expect.

[Chris] tried the community-supported PPC32 port of Debian Sid, but the installer didn’t work reliably, and driver issues made running it “Death by a thousand cuts”. NetBSD, with it’s institutional obsession with running on anything and everything, works perfectly on this legally-adult hardware. Even better, [Chris] reports NetBSD running considerably faster, getting 60 FPS in pygame vs 25 FPS under Linux.

This is almost certainly not the year of the BSD Desktop, but if you’ve got an old PPC machine you feel like dusting off to enjoy a low-powered modern workflow, NetBSD may be your AI-code-free jam. It’s great to see old hardware still doing real work. If you’d rather relive the glory days, you can plug that PPC into a wayback proxy to browse like it’s 2005 again. If you get bored of nostalgia, there’s always MorphOS, which still targets PPC.

hackaday.com/2025/07/23/game-d…

One of the earliest commercially-successful camera technologies was the rangefinder — a rather mechanically-complex system that allows a photographer to focus by triangulating a subject, often in a dedicated focusing window, and and frame the shot with another window, all without ever actually looking through the lens. Rangefinder photographers will give you any number of reasons why their camera is just better than the others — it’s faster to use, the focusing is more accurate, the camera is lighter — but in today’s era of lightweight mirrorless digitals, all of these arguments sound like vinyl aficionados saying “The sound is just more round, man. Digital recordings are all square.” (This is being written by somebody who shoots with a rangefinder and listens to vinyl).

While there are loads of analog rangefinders floating around eBay, the trouble nowadays is that digital rangefinders are rare, and all but impossible to find for a reasonable price. Rather than complaining on Reddit after getting fed up with the lack of affordable options, [Mr.50mm] decided to do something about it, and build his own digital rangefinder for less than $250.

Part of the problem is that, aside from a few exceptions, the only digital rangefinders have been manufactured by Leica, a German company often touted as the Holy Grail of photography. Whether you agree with the hype or consider them overrated toys, they’re sure expensive. Even in the used market, you’d be hard-pressed to find an older model for less than $2,000, and the newest models can be upwards of $10,000.

Rather than start from scratch, he fused two low-cost and commonly-available cameras into one with some careful surgery and 3D printing. The digital bits came from a Panasonic GF3, a 12 MP camera that can be had for around $120, and the rangefinder system from an old Soviet camera called the Fed 5, which you can get for less than $50 if you’re lucky. The Fed 5 also conveniently worked with Leica Thread Mount (LTM) lenses, a precursor to the modern bayonet-mount lenses, so [Mr.50mm] lifted the lens mounting hardware from it as well.

Even LTM lenses are relatively cheap, as they’re not compatible with modern Leicas. Anyone who’s dabbled in building or repairing cameras will tell you that there’s loads of precision involved. If the image sensor, or film plane, offset is off by the slightest bit, you’ll never achieve a sharp focus — and that’s just one of many aspects that need to be just right. [Mr.50mm]’s attention to detail really paid off, as the sample images (which you can see in the video below) look fantastic.

With photography becoming a more expensive hobby every day, it’s great to see some efforts to build accessible and open-source cameras, and this project joins the ranks of the Pieca and this Super 8 retrofit. Maybe we’ll even see Leica-signed encrypted metadata in a future version, as it’s so easy to spoof.

youtube.com/embed/uzDwgYQVdWs?…

hackaday.com/2025/07/23/fusing…