OpenAI focuses on growth, not IPO - Sync #544

Plus: Anthropic projects $70B in revenue by 2028; Apple to use Google AI for $1B; Microsoft's future without OpenAI; signs of introspection in AI models; real-life artificial wombs; and more!

Hello and welcome to Sync #544!

This week, OpenAI’s top executives shared new insights into the company’s finances, potential IPO, and future plans—which we’ll take a closer look at.

Elsewhere in AI, Anthropic projects $70 billion in revenue by 2028, Apple is nearing a $1 billion-per-year deal to use Google’s AI models for Siri, and Microsoft has outlined its vision for a future without OpenAI. Additionally, China has banned foreign AI chips from state-funded data centres, AI musicians are making their way onto the music charts, and researchers at Anthropic have found signs of introspection in modern AI models.

Over in robotics, Xpeng has unveiled its humanoid robot, Elon Musk has announced that production of the Cybercab will begin in April next year, and Anduril has flown its uncrewed jet drone for the first time.

Elsewhere, this week’s issue of Sync also features real-life artificial wombs, a new human brain model aiding the study of brain diseases, Musk’s $1 trillion pay package, insights on what skills engineers should develop in the age of AI, and more!

Enjoy!

OpenAI focuses on growth, not IPO

OpenAI says IPO not imminent as record spending and data centre deals raise questions

OpenAI has been at the centre of many multibillion-dollar deals this year as it races to build the infrastructure needed for serving AI at a massive scale. In recent weeks, the company has signed a $300 billion deal with Oracle, announced a collaboration with Broadcom to deploy 10 gigawatts of OpenAI-designed AI accelerators, and formed a partnership with AMD to roll out 6 gigawatts of GPUs. In September, OpenAI struck a $100 billion agreement with Nvidia to deploy 10 gigawatts of its AI systems. This week, OpenAI signed a seven-year, $38 billion deal with Amazon to secure cloud computing power. The Amazon agreement, OpenAI’s first with the company, will give it access to additional computing power it needs to train AI models and run ChatGPT.

The sheer scale of these deals has raised questions about OpenAI’s finances and long-term funding strategy. Currently, OpenAI has no obligation to disclose its financial results to the public. But that could change in the next few years. Since the company completed its transition to a for-profit structure, rumours about a possible IPO have circulated. Reuters has reported that OpenAI may be laying the groundwork for a $1 trillion listing, the largest IPO in history, that could happen as soon as 2027 or even late 2026.

However, those expecting it to happen soon will have to wait. OpenAI’s Chief Financial Officer, Sarah Friar, said that an IPO is “not on the cards” in the near term. Speaking at The Wall Street Journal’s Tech Live conference, Friar said the company is prioritising growth and R&D over profitability. “We are continuing to get the company into a state of constantly stepping up into the scale we are at, so I don’t want to get wrapped around an IPO axle,” she said.

Friar said OpenAI could reach profitability quickly if it chose to reduce investment, but that the company’s focus remains on expansion. “I’m not overly focused on a break-even moment today,” she said. “I know if I had to get to break-even, I have a healthy enough margin structure that I could do that by pulling back on investment.” She also revealed that OpenAI expects to generate about $13 billion in revenue this year, with enterprise sales now making up roughly 40% of total revenue, up from 30% at the start of the year.

Friar’s comments about potentially seeking government support for data-centre financing led to speculation that OpenAI might be looking for public guarantees to back its infrastructure spending. Sam Altman moved quickly to clarify the issue. In a lengthy post on X, he wrote: “We do not have or want government guarantees for OpenAI datacentres. We believe that governments should not pick winners or losers, and that taxpayers should not bail out companies that make bad business decisions.”

Altman said the company’s only discussions about loan guarantees involved semiconductor manufacturing, not AI data centres, and that those conversations were focused on rebuilding the US chip supply chain. He reiterated that OpenAI plans to fund its massive infrastructure push through continued growth, equity sales, and debt, not government intervention.

Altman also used the opportunity to outline OpenAI’s longer-term ambitions. He said the company expects to exceed $20 billion in annualised revenue run rate by the end of this year and to grow to “hundreds of billions” by 2030. Beyond ChatGPT subscriptions, OpenAI is exploring new business lines—including enterprise software, consumer AI devices, robotics, and scientific discovery tools—as well as plans to sell computing power directly as an “AI cloud” provider.

Altman’s comments suggest that OpenAI is on track to surpass its projected 2025 revenue goal of $12.7 billion. However, there is still no indication that the company is close to profitability. OpenAI reportedly generated $4.3 billion in revenue during the first half of 2025 but recorded losses of around $13.5 billion over the same period. The company itself does not expect to reach profitability until 2029. For now, OpenAI’s focus is on expansion. As Altman put it, “We are trying to build the infrastructure for a future economy powered by AI.”

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

---

🦾 More than a human

This machine could keep a baby alive outside the womb. How will the world decide to use it?
Artificial wombs sound like something straight out of a sci-fi story, but there are startups that are working on making them a reality. The AquaWomb, developed by a Dutch team, aims to support extremely premature babies—those born around 22 to 24 weeks—by recreating the protective environment of the womb through a fluid-filled tank and artificial placenta. While the technology could save countless lives, it also raises deep ethical, emotional and legal questions about what it means to nurture life outside the human body.

🧠 Artificial Intelligence

Microsoft Lays Out Ambitious AI Vision, Free From OpenAI
Microsoft’s AI Chief Executive, Mustafa Suleyman, has unveiled a new vision for the company’s artificial intelligence strategy, focusing on developing models with superintelligence while reducing reliance on OpenAI. The company’s new MAI Superintelligence Team will focus on creating powerful but safe AI that stays under human control and supports useful goals like improving healthcare, science, and clean energy. While Microsoft will continue working with OpenAI, it is also building its own AI models and infrastructure to gain more independence in developing advanced AI.

Microsoft signs $9.7 billion cloud deal with IREN as AI demand swells
Microsoft has signed a $9.7 billion agreement with data-centre operator IREN to secure access to Nvidia’s advanced chips. The five-year partnership lets Microsoft boost capacity without building new data centres or spending heavily on rapidly ageing chips. IREN will use $5.8 billion of the funds for Nvidia GB300 chips and Dell equipment, with deployment set through 2026 at its Texas site.

Anthropic projects $70B in revenue by 2028
According to The Information, Anthropic is expecting to bring in up to $70 billion in revenue and $17 billion in cash flow by 2028. The company aims to reach $9 billion in annual recurring revenue this year and up to $26 billion in 2026. Its API sales are projected to hit $3.8 billion this year, twice that of OpenAI. Additionally, Anthropic expects its gross profit margin to reach 50% this year and 77% in 2028, up from negative 94% last year.

Meta’s $75B AI Infrastructure Bet: Inside the Biggest Cloud Deals of 2025
Meta, just like any other tech giant, is heavily investing in AI infrastructure. In just three months, the company has announced $75 billion, or a third of its Q3 2025 revenue, in infrastructure deals. This post dives deep into unpacking those deals, the infrastructure Meta is building, and who its key partners are in this massive project.

Apple Nears $1 Billion-a Year Deal to Use Google AI for Siri
Apple is close to a $1 billion-a-year deal with Google to use its Gemini AI model to upgrade Siri, Bloomberg’s Mark Gurman reports. The partnership will help Apple rebuild Siri’s technology and add smarter features, while keeping user data on Apple’s own secure servers. Although Apple plans to develop its own AI models in the future, it’s relying on Google’s system for now to speed up progress. The new version of Siri is expected to launch next spring.

Google’s rolling out its most powerful AI chip, taking aim at Nvidia with custom silicon
Google has announced the public release of its most advanced AI chip yet, the seventh-generation Tensor Processing Unit (TPU) named Ironwood. Built in-house, Ironwood is designed for tasks ranging from large model training to real-time AI applications, with the ability to link over 9,000 chips in a single pod to remove data bottlenecks. Alongside Ironwood, Google is upgrading its cloud services to compete with Amazon and Microsoft and has raised its spending plans to $93 billion to meet growing demand for AI technology.

China bans foreign AI chips from state-funded data centres
China has ordered all state-funded data centre projects to use only domestically made artificial intelligence chips, Reuters reports. Projects that are less than 30% complete must remove or cancel orders for foreign chips like those from Nvidia, AMD, and Intel, while others will be reviewed on a case-by-case basis. The move aims to reduce China’s reliance on foreign technology and boost its own chip industry, but it could hurt US companies such as Nvidia, which has already lost most of its market share in China.

Shopify says AI traffic is up 7x since January, AI-driven orders are up 11x
Shopify says artificial intelligence is now central to its business, calling it the biggest tech shift since the internet. The company, which recently partnered with OpenAI, has seen AI-driven traffic to its stores rise sevenfold and AI-powered purchases grow elevenfold this year. Shopify is also working with Perplexity and Microsoft Copilot to bring shopping into AI chats as it builds toward “agentic commerce.”

Google Gemini’s Deep Research can look into your emails, drive, and chats
Google’s Gemini Deep Research can now use your emails, documents, and chats to create more detailed reports. It combines information from Gmail, Drive, Chat, and the web to help with tasks like market research or competitor analysis. Users can choose which sources Gemini accesses and export results as documents or podcasts. The feature is available on desktop now and will reach mobile devices soon.

Billboard Says AI-Powered ‘Artists’ Are Increasingly Hitting The Charts
Xania Monet appeared seemingly out of nowhere, achieving multiple Billboard chart placements, including a No. 1 R&B Song Sales hit, and later signing a multimillion-dollar deal with Hallwood Media. Xania, however, is not human but part of a growing wave of AI artists such as Breaking Rust, Juno Skye, and Enlly Blue, who have gained popularity and large online followings. These AI musicians have also sparked debate about copyright, fairness, and creativity, with critics arguing they use other artists’ work to train their models and lack genuine human expression.

Signs of introspection in large language models
Researchers at Anthropic have released a paper presenting evidence that current Claude models display some degree of introspective awareness and limited control over their own internal states. However, they emphasise that this introspective ability is highly unreliable and narrow in scope, and there is no indication that AI models introspect in the same way humans do. If models could reliably introspect, it could lead to improvements in transparency, debugging, and the trustworthiness of AI systems.

AI Will Be Your Personal Political Proxy
The authors of this article propose that AI could act as political proxies—digital agents that understand citizens’ preferences and can vote or advocate on their behalf. They argue this could make democracy more accurate, responsive, and continuous. However, they also warn of serious risks, such as bias, misrepresentation, and the loss of human involvement and civic engagement, stressing that these digital proxies must be designed to support—not replace—human involvement in democracy.

🤖 Robotics

Xpeng founder unzips humanoid robot to prove it’s not human
Chinese carmaker Xpeng unveiled Iron, its new humanoid robot. Apparently, it looked so real that people thought it was human. To prove otherwise, founder He Xiaopeng unzipped the robot’s back to show its mechanical parts. Xpeng plans to start mass production next year, with steelmaker Baosteel as its first customer.

US defense company Anduril flies its uncrewed jet drone for first time
Anduril’s new jet-powered drone, the YFQ-44A, completed its first flight in California, marking a major step in the company’s work with the US Air Force. The drone flew semi-autonomously, able to control its own flight and land with minimal human input. It is part of the Air Force’s program to develop “loyal wingman” drones that can fly alongside fighter jets, where Andruil competes with other companies, such as General Atomics and Shield AI. Anduril plans to start building the prototype next year.

Tesla to begin Cybercab production in April, Musk claims
Elon Musk announced that production of the Cybercab, a fully autonomous electric vehicle with no pedals, steering wheel, or mirrors, will start in April at Tesla’s Austin, Texas factory. Musk said the new production line could build up to 3 million Cybercabs a year, but regulators still need to approve the car, and there are conflicting reports from Tesla about whether it will include manual controls.

Humanoid Robotics Startup K-Scale to Shut Down After Exploring Sale to 1X, Bot Co.
K-Scale Labs, a Bay Area startup making open-source humanoid robots, is closing down after failing to raise $10-15 million, The Information reports. CEO Ben Bolte said they will refund unfulfilled pre-orders for the $8,000 K-Bot robot. Despite getting $1 million in orders and interest in their open-source approach, attempts to sell the company to companies like 1X and The Bot Co. were unsuccessful.

Waymo’s robotaxi expansion accelerates with 3 new cities
Waymo, Alphabet’s self-driving car company, is launching robotaxi services in Detroit, Las Vegas, and San Diego as it expands rapidly across the US. The move is part of plans to reach one million rides a week by 2026. Waymo will start with human drivers mapping routes before shifting to fully driverless cars.

GEN-0 - Embodied Foundation Models That Scale with Physical Interaction
Generalist AI introduces GEN-0, a new class of embodied foundation models trained directly on large-scale, high-fidelity physical interaction data rather than text or simulation. It combines seeing, thinking, and acting in one continuous process and improves as it gets more data and computing power, with a big jump in ability around seven billion parameters, according to Generalist AI. GEN-0 has been trained on over 270,000 hours of robot experience in different environments and can work across many robot types and tasks.

Researchers at Beijing Academy of Artificial Intelligence (BAAI) trained a Unitree G1 to pull a 1,400 kg car
It is exactly what the title says.

🧬 Biotechnology

MIT researchers invent new human brain model to enable disease research, drug discovery
MIT researchers have created “Multicellular Integrated Brains” (miBrains) — tiny 3D models grown from stem cells that include all six major types of brain cells. These models closely replicate how the human brain functions and can be personalised to study neurological diseases and test treatments. Using miBrains, scientists have already discovered how the APOE4 gene variant linked to Alzheimer’s alters cell interactions to cause the disease.

Caribou results suggest renewed promise for ‘off-the-shelf’ cancer cell therapy
Caribou Biosciences’ “off-the-shelf” cancer therapy, vispacabtagene regedleucel (vispa-cel), produced responses in over 80% of lymphoma patients in an early trial, matching results from personalised CAR-T treatments. Using donor cells matched by immune proteins (HLAs), it showed lasting remissions, mild side effects, and potential for outpatient use. The promising results support a larger Phase 3 trial and suggest vispa-cel could be a simpler, more accessible alternative to existing therapies.

💡Tangents

Tesla Shareholders Approve Elon Musk’s $1 Trillion Pay Package
Tesla shareholders have approved an unprecedented $1 trillion pay package for Elon Musk, contingent on the company meeting ambitious goals by 2035, including reaching an $8.5 trillion valuation, producing 20 million vehicles, and deploying 1 million robots. The approval, supported by over 75% of shareholders, could increase Musk’s stake in Tesla from 12% to 25%. The move follows legal challenges over his earlier $50 billion pay package and is meant to keep him focused on leading Tesla’s work in electric cars, robotics, and self-driving technology.

▶️ Hardware Skills for the Age of AI (12:18)

Every engineer will at some point face a dilemma—specialise in cutting-edge technology and risk vulnerability when markets shift, or stick to traditional work and miss the field’s most exciting developments. In this video, Leon Zhang from Leon Ex Machina proposes a third path—developing high-value, transferable skills through strategic side projects. Skills that matter whether that shiny new technology goes boom or bust. He then gives solid advice on how to walk the third path that can be applied to any kind of engineering, not just mechanical or robotics.

Thanks for reading. If you enjoyed this post, please click the ❤️ button and 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!"