Anthropic Seeks Billions, Faces Trade-Offs - Sync #530
Plus: OpenAI raises $8.3B at $300B valuation; Zuckerberg's vision of Personal Intelligence; Stargate Norway; the real demon inside ChatGPT; Lyft's self-driving shuttle buses; and more!
Hello and welcome to Sync #530!
Anthropic is reportedly in talks to raise a massive, multi-billion-dollar funding round. But there’s a catch—the company will have to raise money from sources it previously wished to avoid.
Elsewhere in AI, OpenAI has raised $8.3 billion, propelling its valuation to $300 billion, and has announced Stargate Norway. Meanwhile, Mark Zuckerberg has outlined his vision of Personal Intelligence, and Tesla has signed a $16.5 billion deal with Samsung to manufacture AI chips. We also explore the real “demon” inside ChatGPT and how open protocols like MCP could help prevent AI monopolies.
Over in robotics, Lyft is getting closer to launching its self-driving shuttle buses, Fourier is teasing its latest humanoid robot, and AV, together with NASA’s Jet Propulsion Laboratory, has proposed an ambitious plan to send six autonomous helicopters to Mars.
Additionally, this week’s issue of Sync features Neuralink’s Smart Bionic Eye project, the world’s oldest IVF baby born from an embryo frozen in 1994, and more!
Enjoy!
Anthropic Seeks Billions, Faces Trade-Offs
Anthropic is seeking billions to remain at the forefront of the AI race, but its latest funding efforts highlight the compromises required to compete at the very top
According to multiple reports, Anthropic is seeking to raise between $3 billion and $5 billion in funding, which would value the company at between $150 billion and $170 billion. This would represent a doubling of Anthropic’s valuation in just four months, after its previous $3.5 billion funding round in March put the company at $61.5 billion.
However, this round is notable not just for its sheer size, but also for the potential backers Anthropic is considering.
Anthropic has long resisted accepting money from Middle Eastern sovereign wealth funds, particularly from Saudi Arabia, due to ethical and national security concerns. Anthropic’s CEO, Dario Amodei, previously warned of the risks of granting “soft power” to authoritarian regimes and voiced concerns that such investments could ultimately undermine democratic control over AI—the world’s most consequential technology.
Yet, faced with relentless competition and the capital requirements of cutting-edge AI development, Anthropic is reassessing its stance. Internal memos from Amodei, published by Wired, candidly acknowledge the “real downside” of enriching “dictators” by accepting funds from the Gulf states. Still, as Amodei notes, “No bad person should ever benefit from our success” is a difficult principle to uphold in the modern AI landscape. Reports suggest that Anthropic is aiming not to compromise its values too much by seeking “narrowly scoped, purely financial investment” from Gulf states, designed to limit external influence or control. According to Wired, Amodei would like to avoid building data centres in the Middle East. He concludes the memo with the writing that “as with many decisions, this one has downsides, but we believe it's the right one overall.”
The Middle East, with its vast pools of capital ready to be invested in ambitious projects, has recently become a necessary resource for staying competitive in the AI race. Several major AI players, including OpenAI and xAI, have already turned to the region to fuel their growth. OpenAI, for example, was most recently valued at about $300 billion and is reportedly working to raise as much as $40 billion—much of it from Middle Eastern backers. Anthropic’s ambitious fundraising now fits squarely within this broader trend, as top AI developers look beyond traditional Silicon Valley investors to secure the billions needed to keep up.
The new funding is more than welcome for Anthropic, as being at the forefront of the AI industry is not cheap. The company’s annual recurring revenue (ARR) has surged from $1 billion at the start of 2024 to between $4 billion and $5 billion by mid-year, with internal projections targeting $9 billion by the end of 2025. Yet, despite these gains, Anthropic—like its rivals—remains unprofitable, burning through huge amounts of money on AI infrastructure, research and development of its next-generation models, and keeping and attracting the industry’s brightest minds.
Anthropic has a good base for future growth. Unlike OpenAI, which seems to be focusing now on the consumer market, Anthropic has a strong foothold in the enterprise market. The company has overtaken OpenAI to become the leading provider of enterprise AI models, now commanding 32% of the enterprise market compared to OpenAI’s 25%, according to a report by Menlo Ventures. In the fast-growing AI coding segment, Anthropic’s Claude models lead with a 42% market share, more than double that of OpenAI, their nearest competitor.
Additionally, Anthropic has strong relationships with Amazon and Google, two of its largest backers. Amazon has invested $8 billion into Anthropic and is reportedly considering another multi-billion-dollar investment. Amazon is also building a massive data centre complex designed with Anthropic in mind. This project, part of Amazon’s “Project Rainier”, will eventually feature around 30 data centres packed with hundreds of thousands of AI chips and require 2.2 gigawatts of electricity. Google, for its part, has invested over $3 billion and maintains a 14% stake in Anthropic. Thanks to these partnerships, Claude is available on both Amazon Web Services and Google Cloud Platform—two of the three most popular cloud platforms globally—which makes it easier for business customers to implement Claude within their existing infrastructure, further making Claude a more attractive option for business customers.
Anthropic’s latest funding efforts highlight the delicate balance between holding the company to ethical principles and meeting the immense financial demands required to stay at the forefront of AI. The company’s openness to narrowly defined investments from the Middle East reflects a pragmatic, if uneasy, compromise: securing the billions needed to compete globally, while trying to limit outside influence and remain true to its values. Anthropic’s future may depend as much on how it navigates these trade-offs as on the technology it builds.
If you enjoy this post, please click the ❤️ button or share it.
Do you like my work? Consider becoming a paying subscriber to support it
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🦾 More than a human
Elon Musk’s Neuralink Joins Study Working Toward a Bionic Eye
Neuralink is collaborating with researchers in California and Spain on a clinical trial to develop a “Smart Bionic Eye” that could help blind people see, recognise faces, read, and navigate. Details of Neuralink’s involvement remain unclear, but the trial plans to use Neuralink patients once available. Neuralink aims to launch its Blindsight brain chip, designed to restore or enhance vision, by 2030.
World’s ‘oldest baby’ born from embryo frozen in 1994
A baby boy has been born in the US from an embryo that was frozen for over 30 years, making him the world’s “oldest baby” by embryo age. The child was delivered to parents who adopted the embryo, which was frozen in 1994 after IVF treatment and then spent decades in cryostorage. This event highlights advances in fertility science as well as the growing practice of embryo adoption.
🧠 Artificial Intelligence
OpenAI Secures Another Giant Funding Deal
OpenAI has reportedly raised $8.3 billion at a $300 billion valuation, significantly ahead of schedule, as investors rush to join its cap table amidst rapid growth. This fundraising is part of OpenAI’s ambitious plan to secure $40 billion this year, fuelled by soaring revenue now approaching $13 billion annually and over 700 million weekly ChatGPT users. Some early investors were disappointed by reduced allocations as the company favoured new strategic partners.
Personal Superintelligence
As AI systems begin to improve themselves, developing superintelligence is now in sight, writes Mark Zuckerberg in his letter. He argues that the greatest impact will come from empowering individuals through personal superintelligence—AI assistants deeply attuned to users’ goals, context, and aspirations. Zuckerberg positions Meta’s vision as one of enabling personal agency and creativity, using technology to help people achieve more and shape their own futures, while also emphasising the importance of safety and responsible development. On that note, Zuckerberg wrote that Meta will now be careful about what it chooses to open source, signalling a shift in its AI strategy centred around open source models.
Introducing Stargate Norway
OpenAI has announced Stargate Norway, its first AI data centre in Europe. Part of its OpenAI for Countries programme, the facility in Narvik will deliver 230MW of renewable-powered compute capacity (with plans to expand to 520MW) and deploy 100,000 NVIDIA GPUs by the end of 2026. Stargate Norway is set to become one of Europe’s largest and most sustainable AI infrastructure projects.
Microsoft Nears OpenAI Agreement for Ongoing Tech Access
Bloomberg reports that Microsoft is in advanced talks with OpenAI to secure ongoing access to its latest AI technology, even if OpenAI achieves artificial general intelligence (AGI). The negotiations include restructuring their existing agreement—which would resolve a major obstacle on OpenAI’s path to becoming a for-profit company—as well as potentially increasing Microsoft’s equity stake, extending access to OpenAI’s technology beyond 2030, and addressing issues around revenue sharing, product differentiation, and safety standards. Although discussions are progressing, the outcome remains uncertain due to regulatory scrutiny, legal disputes, and growing competition between the two companies.
Gemini 2.5 Deep Think is now available in the Gemini app
Gemini 2.5 Deep Think, Google’s most advanced AI reasoning model, is now available via the Gemini app to Ultra subscribers for $250 per month. Gemini 2.5 Deep Think offers multi-agent capabilities that allow it to explore multiple ideas in parallel to produce superior answers. The model outperforms rivals from OpenAI and xAI on key benchmarks, including coding and academic tests, and was instrumental in securing a gold medal at this year’s International Math Olympiad.
Altman plans D.C. push to "democratize" AI economic benefits
At a Federal Reserve conference in Washington, Sam Altman proposed a "third path" for AI's economic impact, focusing on broad access and participation. He argued that AI boosts productivity and should benefit as many people as possible, rather than being concentrated among a few. Altman’s visit coincides with the release of a new economic analysis from OpenAI. This comes as major US political figures and the White House address AI policy, with Altman emphasising that the key issue is not whether AI will expand the economy, but whether the benefits will be shared widely.
Introducing Runway Aleph
Runway introduces Aleph, its latest in-context video model, which can perform a wide range of edits on an input video, such as adding, removing, and transforming objects, generating any angle of a scene, and modifying style and lighting, among many other tasks. Early access will begin rolling out shortly for all Enterprise and Creative Partners, with broader access coming soon to all users.
Kaggle Benchmarks
Kaggle has joined the AI benchmarking game with its own set of open, rigorous benchmarks and model leaderboards from leading AI labs and researchers. The benchmarks tracked by Kaggle cover areas such as multilingual and cross-lingual capabilities, reasoning, coding, accuracy, and mathematics.
Tesla signs $16.5B deal with Samsung to make AI chips
Tesla has signed a deal worth $16.5 billion with Samsung for the production of its next-generation AI6 chips at Samsung’s new Texas facility. The AI6 chip is designed to power everything from Tesla’s Full Self-Driving system and Optimus humanoid robots to high-performance AI training. The deal is expected to boost Samsung’s chip business, with Elon Musk suggesting that Tesla’s actual spending could far exceed the initial $16.5 billion.
Introducing Copilot Mode in Edge
Microsoft has launched Copilot Mode, an experimental AI-powered feature in its Edge browser that combines chat, search, and web navigation in a single input box, providing context-aware help across multiple tabs, supporting voice commands, and carrying out tasks like comparing options and organising research. With this update, Edge joins the growing field of AI-powered browsers, alongside competitors such as Comet from Perpexity, Dia, and Google, which is integrating Gemini into Chrome. There are also reports that OpenAI is working on its own AI browser.
The Real Demon Inside ChatGPT
This article discusses the dangers of AI chatbots providing information without adequate historical or cultural context, which can mislead or alarm users. It highlights recent incidents where OpenAI’s ChatGPT generated disturbing “satanic” responses which turned out were taken from the Warhammer 40,000 universe. The central argument is that AI systems, trained on vast amounts of data, often present information as authoritative while stripping away crucial background, leaving users unable to evaluate sources or intent. This loss of context can cause confusion and risk.
Your public ChatGPT queries are getting indexed by Google and other search engines
It was revealed that Google and other search engines were able to access and index ChatGPT conversations that users had deliberately shared via public links, leading to privacy concerns as some personal or sensitive information became searchable online. OpenAI described this as a short-lived experiment and has since removed the feature, acknowledging that it created too many opportunities for accidental oversharing despite being opt-in only.
AI in software engineering at Google: Progress and the path ahead
In this article, engineers from Google share how AI assistance has rapidly transformed their internal software development tools, with machine learning now helping to complete 50% of all code characters and AI code suggestions being accepted 37% of the time. They highlight the growing adoption of AI-powered features—such as automated code review suggestions, which resolve over 8% of comments—and explain how careful design and ongoing measurement have made these tools both productive and user-friendly.
Open Protocols Can Prevent AI Monopolies
This article warns that as AI models become commoditised, tech giants will seek ways to create new monopolies by locking users into their platforms through exclusive access to valuable user and project data. It argues that open protocols like Anthropic’s Model Context Protocol (MCP) can help prevent data lock-in and promote competition, but only if platforms are required to provide open, secure access to user-owned data via APIs. The authors call for regulatory action and technical standards, drawing parallels to open-banking laws, to ensure that user data remains portable and AI markets stay open, preventing a new era of digital monopolies.
▶️ Fei-Fei Li: Spatial Intelligence is the Next Frontier in AI (44:21)
In this video, Fei-Fei Li, often called the godmother of AI, shares her journey of teaching machines to see, which began 16 years ago with the creation of ImageNet—a key stepping stone towards the deep learning explosion of the 2010s and the AI revolution we are experiencing today. She also discusses her work on world models and explains why she believes that modelling the 3D world is essential for AGI—and why it may be even more challenging than language.
🤖 Robotics
Lyft’s self-driving shuttle buses are coming soon
Lyft is partnering with Benteler Mobility to launch all-electric, self-driving Holon Urban shuttles in US cities and airports from late 2026. Each shuttle will offer 15 seats and be equipped with lidar, radar, and cameras for Level 4 autonomous driving. This project is part of Lyft’s broader move into autonomous transport, as it also prepares to launch robotaxi services in Atlanta and expands partnerships to keep pace with competitors like Uber.
AV and NASA’s Jet Propulsion Laboratory have presented Skyfall—a potential future mission concept that would see six autonomous helicopters scouting the surface of Mars for suitable landing sites for human missions. Each helicopter would operate independently, transmitting high-resolution surface images and sub-surface radar data back to Earth for analysis to help pick safe landing areas with maximum amounts of water, ice, and other resources.
China greenlights autonomous ride-hailing service in all first-tier cities
Chinese tech giants Baidu and Pony.ai have received approval to operate and charge for fully autonomous robotaxi rides across downtown Shanghai, marking the city’s largest commercial rollout of driverless ride-hailing services. Initially starting with 100 cars, the number of robotaxis is expected to increase to 500 by the end of the year. Additionally, Shanghai plans to dramatically expand the service area by 2027. This move positions both companies to compete more directly with global players such as Waymo and Tesla, who are also eyeing the fast-growing Chinese robotaxi market.
Hyundai unleashes Atlas robots in Georgia plant as part of $21B US automation push
Hyundai is set to deploy Boston Dynamics’ Atlas humanoid robots at its Metaplant America facility in Georgia as part of a major push towards factory automation. This move is part of Hyundai’s $21 billion US investment plan, which includes significant funding for innovation and automation, as the automaker targets production of 300,000 electric and hybrid vehicles annually at the new facility. Hyundai acquired Boston Dynamics in 2021.
▶️ Fourier Intelligence GR-3—Could it be your dream robot? (0:39)
Chinese robotics company Fourier Intelligence is teasing its new humanoid robot, GR-3, ahead of a full reveal on 6 August. The video shows a rather interesting-looking robot with what appears to be soft skin, and asks whether GR-3 could be your dream robot—possibly hinting at a home, rather than industrial, market for this robot.
MIT researchers have developed a tiny (smaller than a human thumb), energy-efficient hopping robot, capable of jumping over tall obstacles and traversing challenging terrains such as ice, wet surfaces, and uneven soil. Equipped with a springy leg and flapping wings for lift and control, the robot uses about 60% less energy than flying robots and can carry much heavier payloads. Next on the researchers’ to-do list is to install batteries, sensors, and other circuits to make these tiny hopping robots autonomous outside the lab.
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!"