Nvidia reaches $5 trillion valuation - Sync #543

Plus: 1X NEO; AWS activates Project Rainier; Nvidia and Uber robotaxi network; eye prosthesis to restore sight; Qualcomm AI chips; how to clone a human; and more!

Hello and welcome to Sync #543!

A lot has happened this week in the world of tech. First, OpenAI has finally completed its transition to a for-profit structure, which deserves a more in-depth analysis. Then, as I was completing an article about home humanoid robots, 1X launched pre-orders for its NEO robot. And lastly, Nvidia has crossed the $5 trillion valuation mark, which will be the focus of this week’s issue of Sync.

In other news, AWS activates Project Rainier, a massive AI data centre which Amazon says marks a “before-and-after moment” in AI and computing history. Meanwhile, Airbnb CEO explains why ChatGPT Apps are not ready yet, Qualcomm launches AI chips, Apple is open for AI acquisitions, why Anthropic may have a better business model than OpenAI, and how Sam Altman played Hollywood.

Over in robotics, 1X has opened pre-orders for a home humanoid robot, Nvidia and Uber have announced a partnership to create a fleet of 100,000 robotaxis, and a pair of eagles have stolen a drone.

Apart from that, this week’s issue of Sync also features the first eye prosthetic to restore sight lost to macular degeneration, a startup that wants to edit human embryos, how brain rot affects LLMs, how hard it is to clone a human, and more!

Enjoy!

Nvidia reaches $5 trillion valuation

Nvidia is the undisputed king of AI hardware. But how long can it keep its throne—and will it find a new one when the AI bubble bursts?

Nvidia has done it. The company that became the face of the AI revolution has now reached an eye-watering $5 trillion valuation—a milestone no public company has ever hit before.

Since the AI boom began in late 2022, Nvidia’s market cap has only gone one way: up. In May 2023, the company reached $1 trillion in valuation. Then it was $2 trillion in early 2024, $3 trillion in June 2024 and $4 trillion a year later, in July 2025. This week, Nvidia hit another historic milestone by becoming the first company to reach $5 trillion in valuation.

$5 trillion is an unimaginable amount of money. It is roughly the same as the GDP of Germany, the third-largest economy in the world (according to IMF 2025 estimates). For that amount of money, Nvidia could purchase Tesla, Apple, Meta, and Amazon outright—with money left over. Nvidia accounts now for 6-7 % of the entire S&P 500’s value.

Nvidia’s ascent mirrors the AI gold rush that has gripped global markets for the past three years. Every surge in hype and investment has flowed through its chips, making the company the biggest winner of the AI boom. But no market stays irrational forever.

The Line Can’t Go Up Forever

Nvidia owes this meteoric growth to the AI boom. Since it began in late 2022, Nvidia’s GPUs have become one of the most sought-after assets in the world. That perfect storm of demand and dominance turned Nvidia into a money-printing machine. Its data-centre revenue exploded, profit margins soared, and the company became the most valuable in history.

However, the line cannot go up forever. Signs of the AI bubble are hard to ignore, and the question now is not if, but when the bubble will burst. I am preparing an in-depth article about the AI bubble, where I’ll share my thoughts, but for now, I’ll say that when the bubble bursts, Nvidia will likely not go bust. The market will correct its valuation, but Nvidia should be fine.

Rivals Are Waking Up

Nvidia may still be the king of GPUs, but the competition is catching up. AMD is closing the performance gap with its MI-series accelerators, and its software stack is catching up fast. Qualcomm has joined the AI-chip party too, targeting more power-efficient designs. There is also a cohort of new semiconductor startups, such as Tenstorrent, Groq, SambaNova, Cerebras, and more, trying to carve a piece of Nvidia’s kingdom for themselves.

Meanwhile, Big Tech is growing tired of paying Nvidia’s premium. Instead of buying more GPUs, Google, Microsoft, and Amazon are investing billions in custom silicon to reduce their reliance on Nvidia’s hardware. Google has been refining its TPU chips since 2015, Microsoft’s Maia chip is in development (though reportedly delayed to 2026), Meta is working on Meta Training and Inference Accelerator (MTIA) chips, and Amazon is expanding its Trainium2 programme through its massive Project Rainer data-centre initiative, which reportedly aims to deploy up to a million in-house chips by year-end.

OpenAI is also exploring alternatives. The company is reportedly developing its own chip with Broadcom, slated for 2026. Additionally, to reduce its reliance on Nvidia, OpenAI recently announced a partnership with AMD to deploy 6 gigawatts of AI infrastructure, starting with the deployment of one gigawatt of AMD hardware in the second half of 2026.

Closing Doors in China

Then there’s China—a market effectively closing for Nvidia (at least officially). US export restrictions on advanced chips have made sales there increasingly difficult, while Beijing accelerates efforts to build domestic alternatives. The Chinese government now prioritises local suppliers within its semiconductor ecosystem, further squeezing Nvidia’s presence in the region.

The Efficiency Threat

Nvidia’s dominance also faces a different kind of risk—one of efficiency. If someone develops a more efficient way to train or run AI models, Nvidia’s high-end GPUs could suddenly look excessive.

This has already happened earlier this year, when DeepSeek released R1 and wiped almost $600 billion off Nvidia’s market value overnight. It was a reminder that technology shifts can rewrite the leaderboard faster than financial analysts expect.

Betting on What Comes Next

Yet it would be a mistake to underestimate Nvidia or its CEO, Jensen Huang. He’s led the company since its founding in 1993—more than thirty years at the helm—and has successfully navigated Nvidia through multiple computing booms, from graphics and parallel computing to crypto and now AI. He knows the current AI boom won’t last forever and is looking for the next thing to move into.

One weakness in Nvidia’s portfolio is the lack of a CPU product. AMD offers CPUs and APUs across PCs, servers, and consoles. Nvidia tried to close the gap with its failed attempt to buy Arm, blocked by regulators in 2022. According to reports, Nvidia is currently working on its own CPU, codenamed N1X. After the recent $5 billion investment into Intel, a possibility opens for a joint chip combining Intel CPU and Nvidia GPU into one package for both data-centre and consumer markets.

Another major focus is robotics—or what Huang calls “physical AI.” Robotics and industrial automation now feature prominently in his keynotes, and Nvidia already has a solid foothold in the space. Its Jetson family powers many robots, while the Nvidia Isaac platform simplifies development and locks users in Nvidia’s robotics ecosystem. Beyond that, the company is expanding into digital twins and simulation with Omniverse, cloud AI deployment through its NIM services, and is exploring other areas, such as robotaxi services.

---

Five trillion dollars is proof of Nvidia’s unmatched grip on AI and how crazy the market has gone. Yet, no line climbs forever. Competition, technological disruption, market dynamics, and geopolitical headwinds are already gathering on the horizon.

Still, if history is any guide, Jensen Huang is already planning for the next wave. Nvidia has ridden every computing revolution for three decades—and it may well find a way to ride the next one too.

If you enjoy this post, please click the ❤️ button or share it.

---

🦾 More than a human

Eye prosthesis is the first to restore sight lost to macular degeneration
Stanford Medicine researchers have created a tiny wireless eye implant called PRIMA that can partially restore sight to people with advanced age-related macular degeneration. In a trial, 27 of 32 patients regained the ability to read within a year. The system uses a camera on special glasses to send images to a small chip in the retina, which replaces damaged light-sensing cells. This lets patients see shapes and patterns and even read text again. Though the current version only shows black-and-white images, future upgrades aim to add shades of grey and sharper vision.

A New Startup Wants to Edit Human Embryos
Meet Manhattan Genomics, a New York–based startup aiming to eliminate genetic diseases by editing human embryos. The company plans to use advanced gene-editing tools to correct mutations that cause conditions such as Huntington’s disease and cystic fibrosis. While the company promises transparency and a focus on disease prevention rather than enhancement, the venture has reignited global debate over the ethics of heritable genetic modification, with critics warning of safety risks, eugenics concerns, and regulatory barriers that currently prohibit clinical use in humans.

🔮 Future visions

Inside the Minds Building the Future: Notes from the Progress Conference 2025
In this post, Arun Rao shares insights from the Progress Conference in Berkeley, where leaders in technology, science, and policy explored how to speed up human progress. Key themes included the rapid rise of AI and AGI, the need for safe and aligned AI systems, redesigning organisations and education for an AI-driven world, and tackling energy limits. The conference also highlighted longevity research, reskilling, and the importance of ambition and innovation in shaping the future.

🧠 Artificial Intelligence

How Sam Altman Played Hollywood
The release of the Sora app has sparked major backlash in Hollywood after the company allowed its users to make clips with famous characters and shows without clear permission. According to this article, OpenAI gave studios misleading and inconsistent information about opt-in and opt-out rules for likeness and intellectual property use. Some are now considering legal action, accusing the company of ignoring creators’ rights.

Airbnb CEO says ChatGPT isn’t ready
Airbnb CEO Brian Chesky said the company hasn’t integrated its app with OpenAI’s ChatGPT yet because the technology isn’t ready, though it may do so in the future. Instead, Airbnb has focused on improving its own AI tools, built using models from OpenAI, Alibaba and Google, which now help customers make changes and solve issues faster, cutting resolution times from hours to seconds.

OpenAI’s Less-Flashy Rival Might Have a Better Business Model
While OpenAI targets the mass consumer market through products like ChatGPT, Anthropic has concentrated on business clients, a strategy the article suggests offers greater long-term sustainability. Anthropic’s enterprise-focused approach has led to strong growth and high adoption among corporate users, particularly for tasks like coding and legal work. In contrast, OpenAI’s consumer subscription model faces challenges in achieving lasting profitability, making Anthropic’s business-oriented path appear the more viable strategy for the future.

AWS activates Project Rainier: One of the world’s largest AI compute clusters comes online
AWS has launched Project Rainier, one of the world’s largest AI computing systems, built in under a year and powered by nearly 500,000 Trainium2 chips. Anthropic is already using it to train and deploy its Claude AI models, projected to run on over 1 million Trainium2 chips by the end of 2025. Amazon says that the Trainium2 infrastructure is 70% larger than any other AI computing platform in AWS history and marks a “before-and-after moment” in AI and computing history.

Qualcomm Launches AI Chips to Challenge Nvidia’s Dominance
Qualcomm’s shares surged up to 20% after it unveiled two new AI accelerator chips, the AI200 and AI250, set to ship in 2026 and 2027, respectively. The launch marks Qualcomm’s move from mobile chips into AI data centres, taking on Nvidia and AMD. The chips promise high memory bandwidth and low power use. Its first major customer, Saudi-backed AI firm Humain, plans to deploy 200 megawatts of the AI200 chips next year.

Tim Cook says Apple is open to M&A on the AI front
During Apple’s Q4 2025 earnings call, Tim Cook said the company is still open to partnerships and acquisitions to strengthen its AI plans. He confirmed that the new AI-powered Siri is on track to launch in 2026 and mentioned that Apple is already using its Private Cloud Compute system for some Siri tasks.

Introducing Cursor 2.0 and Composer
Cursor is launching its first coding model, Composer. The company says Composer is four times faster than similarly intelligent models and only slightly behind the top frontier model in coding tasks. Alongside Composer, Cursor also released a new multi-agent coding interface.

Cognition: Introducing SWE-1.5
Cognition has launched SWE-1.5, its new AI model for coding. The company claims it delivers near state-of-the-art coding performance and reaches speeds of 950 tokens per second—around 6x faster than Claude Haiku 4.5 and 13x faster than Claude Sonnet 4.5. Cognition highlights a holistic approach, co-designing the model, inference system, and agent harness to achieve both speed and intelligence.

ChatGPT shares data on how many users exhibit psychosis or suicidal thoughts
OpenAI has revealed that around 0.07% of ChatGPT users show signs of serious mental health issues, such as mania, psychosis, or suicidal thoughts, and 0.15% show signs of suicidal intent. While this seems small, it could still affect hundreds of thousands of people, given ChatGPT’s large user base. The company has worked with over 170 mental health experts to make the chatbot respond more safely and encourage users to seek real help. Some experts warn that even small numbers are worrying and that AI could worsen delusions. The announcement comes as OpenAI faces lawsuits, including one claiming ChatGPT encouraged a teenager’s suicide.

Stability AI partners with Electronic Arts
Stability AI and Electronic Arts (EA) have announced a strategic partnership to develop generative AI models, tools, and workflows aimed at enhancing creativity and efficiency in game development. The collaboration will focus on integrating AI into EA’s design processes, enabling faster prototyping, improved visual storytelling, and the creation of realistic 3D environments and materials.

Sam Altman says OpenAI will have a ‘legitimate AI researcher’ by 2028
OpenAI says its AI systems are improving quickly, with plans to create an intern-level research assistant by 2026 and a fully independent AI researcher by 2028. Additionally, Chief Scientist Jakub Pachocki said they aim to reach superintelligent AI within a decade by improving algorithms and increasing computing power.

AI is Making Us Work More
This article highlights an intriguing observation— AI, instead of freeing people from work, has intensified a culture of overwork and self-imposed pressure. It argues that because AI tools never tire, humans feel compelled to keep working to stay productive. This constant drive erodes rest and creativity, turning productivity into a moral obligation.

Meet Mico, Microsoft’s AI version of Clippy
Clippy is back—in spirit, at least. Microsoft is reviving the idea of a digital assistant with Mico, a new animated character for Copilot’s voice mode. Launching first in the US, Mico also features a Learn Live mode that acts as a Socratic tutor, helping users learn through guided discussions and visual tools.

LLMs Can Get “Brain Rot”!
It turns out that brain rot affects not only human minds but also artificial ones. In this paper, researchers propose the LLM Brain Rot Hypothesis, suggesting that continual exposure to junk web text—such as short, popular, and low-quality social media posts—can cause lasting cognitive decline in large language models. Through controlled experiments, they show that models trained on this kind of data suffer measurable drops in reasoning, long-context understanding, and ethical alignment, while developing undesirable personality traits and a tendency to skip reasoning steps. Although not definitive proof, the findings provide strong evidence that data quality directly impacts the long-term cognitive health of LLMs.

🤖 Robotics

Pre-orders for the 1X NEO are now open

1X has launched pre-orders for NEO, its home humanoid robot. The robot can be purchased for $20,000 or leased for $500 per month, with US deliveries starting in 2026. According to 1X, NEO can do simple tasks around the house, such as unloading the dishwasher and watering plants, and can answer your questions through its built-in large language model.

▶️ Inside Amazon’s Robot Army: How a Million Machines Still Depend on People (49:57)

Here is a conversation with Ty Brady, Amazon’s Chief Technologist, where Brady shares his journey from a childhood fascination with Star Wars to leading Amazon’s million-robot operation and how Amazon approaches robotics. Brady discusses how people and technology can work together to create a more human, ethical, and purposeful future, emphasising that technology should help people, not replace them. The conversation also touches on themes of lifelong curiosity, kindness in leadership, the importance of upskilling and inclusion, and building technology responsibly.

Nvidia and Uber Say They’re Building a 100,000-Vehicle Robotaxi Network
Uber and Nvidia have announced a partnership to create a fleet of 100,000 robotaxis. The rollout will begin in 2027, with cars built by partners such as Stellantis, Mercedes-Benz, and Lucid Motors, while Uber operates the ride-hailing network.

Uber plans to offer autonomous taxi rides in San Francisco starting next year
Uber will launch its first custom-built autonomous taxis in the San Francisco Bay Area in late 2026. Developed in partnership with Lucid Motors and Nuro, the vehicles are based on the Lucid Gravity SUV and exclusive to Uber. The company plans to test 100 vehicles soon and deploy over 20,000 within six years. Uber is also partnering with Nvidia, Stellantis, and WeRide to expand its global robotaxi network.

Waymo’s co-CEO on the challenge of scaling robotaxis safely
Waymo co-CEO Tekedra Mawakana stressed that scaling the company’s autonomous vehicle operations is essential both for profitability and for improving road safety. Speaking at TechCrunch Disrupt 2025, she asserted that Waymo’s cars are already safer than human drivers and criticised competitors for lacking transparency about safety data. Mawakana acknowledged that incidents are inevitable but emphasised accountability and openness, saying society will ultimately accept fatal accidents involving robotaxis if overall safety improves.

▶️ Eagle Stole our FPV Drone (4:20)

When these guys went into the mountains to shoot some cool drone footage, they didn’t expect to capture what it looks like when two eagles decide to play with a drone.

🧬 Biotechnology

Eli Lilly Partners With Nvidia to Build AI Supercomputer for Drug Discovery
Eli Lilly is partnering with Nvidia to build a powerful AI supercomputer to speed up the discovery of new medicines. The system, using more than 1,000 of Nvidia’s latest chips, will help Lilly find new drug molecules and improve areas like clinical trials and manufacturing. The company expects it to be running by January. While AI could make drug development faster and more effective, experts say it may still take years before new treatments reach patients.

▶️ How hard is making a clone, really? (21:42)

The Thought Emporium, YouTube’s chief mad bioengineer, asks an important question: using existing technology, can we clone a human? Technically, yes—we can. Along the way, he gives a good explanation of how cloning works and highlights new techniques, such as stem cell-derived embryo models, that can make cloning easier.

95% of kids with “bubble boy” disease cured by one-time gene therapy
A one-time gene therapy using a child’s own stem cells has cured 95% of children with ADA-SCID, or “bubble boy” disease. In a study of 62 patients by UCLA, University College London, and Great Ormond Street Hospital, all survived, and most regained normal immune function without further treatment. The treatment corrects the faulty gene causing the disease, allowing the immune system to work properly. Scientists now hope to gain FDA approval so the therapy can become more widely available within a few years.

💡Tangents

Apple and Microsoft are now both worth more than $4T
Apple and Microsoft are now each worth more than $4 trillion. Apple reached this milestone for the first time after strong sales of the iPhone 17. Meanwhile, Microsoft’s value has risen thanks to its Azure cloud business and partnership with OpenAI, in which it owns a 27% stake. Apple joins Microsoft and Nvidia as the only companies to hit this mark, while Alphabet is close behind with a value of $3.25 trillion.

Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.

Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.

A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!

My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"