How Big Tech absorbs AI startups - Weekly News Roundup - Issue #474

Plus: the best bionic leg yet; Apple to get a set on OpenAI's board; Perplexity, Suno and Udio are in trouble; new humanoid robots; Biotech Behemoths in review; and more!

Hello and welcome to Weekly News Roundup Issue #474! We have a packed issue this week, with the main story focusing on the new way Big Tech companies are absorbing smaller companies. We saw this new playbook unfold twice recently with Inflection and Adept, and we can expect it to happen again.

In other news, Apple is reportedly getting a seat on OpenAI’s board. Meanwhile, Perplexity, Suno, and Udio have found themselves in hot water over copyright infringement.

Over in robotics, Figure shares an update on how their humanoid robot is performing in BMW’s factory, and two new humanoid robots join the scene.

Additionally, Neuralink has postponed its second clinical trial, and researchers from MIT have built the best bionic leg yet.

Enjoy!

How Big Tech absorbs AI startups

Tech revolutions (or bubbles, depending on your views) tend to play out roughly in the same way every time.

First, there is a promising idea or technology. A small group of people develop the idea and show its potential. The first startups working on that idea are formed, and the initial rounds of investments are made. As the new big thing gains attention, more people want to join the party. New startups emerge, and more money is poured into these companies in the hope that some will go big. Seemingly in the blink of an eye, a rather large scene has formed.

However, the party cannot last forever. At some point, the winners start to emerge. Maybe they were the first to the party. Maybe they have a better product or a stronger team. Or perhaps they simply have more money and resources than others. In any case, they have something others don’t. The winners begin to capture a larger share of the market and attract more investments. Companies that can’t keep up either close down or get acquired by bigger players.

We saw a similar pattern previously unfolding with the internet companies of the early 2000s and with smartphone and social media companies in the 2010s. We now see the same pattern emerge with all those companies that jumped on the large language model train. However, there is a twist to that pattern that can shape how not only AI but the whole tech industry will operate in the future.

We are entering the market consolidation phase

OpenAI. Anthropic. Inflection. Adept. Mistral. AI21 Labs. Cohere. G42. Aleph Alpha. And many more in China and elsewhere in the world joined the large language model party. Billions have been invested in those companies in recent years to build essentially the same thing—a large language model that can be used in a chatbot like ChatGPT or power third-party apps through APIs.

Over the past 2-3 years, AI companies were in the middle phase. New startups emerged with investment rounds measured sometimes in hundreds of millions and even billions of dollars. Mistral AI, a French AI startup founded by former DeepMind and Meta employees, was founded in 2023 and raised a $113 million seed round for a $260 million valuation while not having any product yet. A year later, Mistral raised an additional €600 million (about $650 million) and reached a valuation of €6 billion ($6.5 billion). Meanwhile, Amazon invested $4 billion into Anthropic and Microsoft put over $10 billion into OpenAI. According to the Stanford AI Index 2024, generative AI startups attracted $25.2 billion in investments last year alone, almost eight times more than in 2022.

However, we now start to see the signs that we are entering the next phase—after the initial explosion comes the phase of acquisitions and consolidation.

The first indication of how eager tech giants are to acquire AI startups came during the OpenAI drama, when Microsoft quickly stepped in and tried to hire Sam Altman and those loyal to him. In the end, OpenAI survived and was not absorbed by Microsoft. Others, however, have not shared the same fate.

The first victims of Big Tech’s new playbook

Inflection AI became the first victim of the consolidation phase. Founded by DeepMind co-founder Mustafa Suleyman, Reid Hoffman, one of the founders of LinkedIn, and Karén Simonyan, a former DeepMind employee, Inflection raised $1.5 billion in total and was valued at $4 billion last year. The company was also building, together with CodeWeave and Nvidia, what was in June 2023 the largest AI cluster in the world to serve its large language model, Inflection-2. The company offered Pi, a personal AI chatbot, and was planning to open access to Inflection-2 through an API for developers to use their large language model.

So it came as a bit of a surprise when news broke out in March 2024 about Microsoft quasi “acqui-hiring” Inflection. Microsoft paid $650 million to license Inflection’s technology and to hire most of the startup’s 70 employees. Among those hired by Microsoft were Mustafa Suleyman and Karén Simonyan, with the former being appointed as the CEO of the newly founded Microsoft AI, Microsoft’s internal AI lab.

What is interesting about that deal is that it had the “hiring” part but not the “acqui” part of “acqui-hiring.” Normally, when a company is acqui-hired, its employees are moved to the acquirer, and the acquired company ceases to exist as an independent entity. In this case, Microsoft did hire most of Inflection’s employees but did not absorb the company. Inflection still exists as an independent company, albeit now resembling a skeleton of what it once was.

A similar fate recently befell another AI startup, Adept. Adept, which raised $415 million in capital and reached a valuation of around $1 billion, is developing an in-house foundation model to power AI agents that can automate software workflows. In May, reports began to emerge that the company was struggling and looking for a buyer. According to The Information, Adept was in talks with Meta and Microsoft regarding a possible acquisition.

In the end, it was not Meta or Microsoft that absorbed the struggling startup—it was Amazon. On June 28th, 2024, Adept announced that its co-founders and some of its team would be joining Amazon. Additionally, Amazon will license Adept’s foundation models, agent technology, and a few datasets. Similarly to what happened with Inflection, whatever is left of Adept still exists as an independent company.

Big Tech’s new and subtle way of absorbing startups

What we have witnessed with Inflection and Adept is a new way for Big Tech to absorb smaller companies, which doesn’t yet have a catchy name. These deals were designed to acquire the talent and technology without acquiring the company itself. The “target” company still exists after the deal is done but only as a shadow of its former self.

One reason for Big Tech to make acquisitions in this way could be to avoid scrutiny from regulators who would take a closer look at these deals if they were traditional acquisitions. This scrutiny could cause delays, extra costs, and even the collapse of the deal. For instance, Amazon’s $1.4 billion deal to acquire iRobot, the creators of Roomba, failed due to the EU blocking the deal. Similarly, Nvidia’s $40 billion plan to acquire ARM failed due to “significant regulatory challenges.” Microsoft, meanwhile, spent 20 months battling with regulators in the US, the UK, and the EU to successfully close the $68.4 billion acquisition of Activision Blizzard.

I don’t think Inflection and Adept will be the only victims of the consolidation phase. I expect this new playbook to unfold again not only in the AI space but possibly in the wider tech space as well. The question now is, who, after Inflection and Adept, will be next?

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

Neuralink Surgery Postponed for Brain-Implant Startup’s Second Patient
Neuralink was set to implant its brain-computer interface into its second patient, but the surgery was halted due to the patient’s medical condition which made them unsuitable to participate in Neuralink’s study of its experimental device. A replacement candidate will likely undergo the surgery next month.

The Best Bionic Leg Yet

Researchers from MIT have created a new type of prosthetic leg that is controlled through neural signals, allowing patients to walk with a natural gait. The approach involves surgery at the amputation site to create what the researchers call an agonist-antagonist myoneural interface (AMI). This creates a two-way connection between body and machine, enabling the prosthetic to move and feel more like a natural leg.

Anti-aging molecule successfully restores multiple markers of youth
Researchers have found that a small molecule called TAC has the potential to regrow neurons, reduce inflammation, and improve cognitive and physical functions in ageing mice. The team screened over 650,000 compounds and identified TAC, which, when administered to elderly mice, led to new neuron growth, better memory, and enhanced muscle strength and coordination. However, more research is needed to confirm TAC’s long-term safety and effectiveness.

▶️ The Last 6 Decades of AI — and What Comes Next | Ray Kurzweil (13:11)

In this TED Talk, Ray Kurzweil reflects on the last 60 years of AI and discusses how AGI can usher in a new age for humanity. He envisions a future where we live longer, healthier lives, solve the major challenges we face today, and become better and smarter beings. And according to Kurzweil, that vision of the future will come much sooner than everyone expects.

🧠 Artificial Intelligence

Apple Poised to Get OpenAI Board Observer Role as Part of AI Pact
Mark Gurman reports in Bloomberg that Apple is set to get a seat on OpenAI’s board as an observer, a position similar to that of Microsoft, OpenAI’s biggest investor and technological partner. According to Gurman, Phil Schiller, the head of Apple’s App Store and its former marketing chief, has been chosen for the role.

Chinese AI firms woo OpenAI users as US company plans API restrictions
Although ChatGPT is not available in China, Chinese users have been able to access OpenAI models through the OpenAI API. However, OpenAI will block access to its API for users in China starting on July 9th. Chinese AI companies see this as an opportunity to attract new users, offering exclusive deals to convince them to use their services over other competitors.

Major record labels sue AI music companies
A group of record labels, including the big three—Universal Music Group (UMG), Sony Music Entertainment, and Warner Records—are suing Udio and Suno, two leading startups in the AI music-generating space, alleging that the companies violated their copyrights “en masse.” The lawsuits were brought by the Recording Industry Association of America (RIAA), the powerful group representing major players in the music industry, along with a group of labels. The RIAA is seeking damages of up to $150,000 per work, along with other fees.

Buzzy AI Search Engine Perplexity Is Directly Ripping Off Content From News Outlets
Perplexity, a startup developing an answer engine to challenge Google Search, has found itself in hot water after Forbes revealed the company is using exclusive content from multiple publications, including Forbes, CNBC, and Bloomberg—many of which are behind a hard paywall—to generate its answers without proper attribution. Perplexity’s CEO, Aravind Srinivas, acknowledged that Perplexity Pages, the feature causing the controversy, has “rough edges” and stated that the feature will improve with time and feedback.

Ever put content on the web? Microsoft says that it's okay for them to steal it because it's 'freeware.’
Mustafa Suleyman, the CEO of the recently created Microsoft AI, said in an interview that all content shared on the web is available to be used for AI training unless a content producer specifically states otherwise. "With respect to content that is already on the open web, the social contract of that content since the 90s has been that it is fair use. Anyone can copy it, recreate with it, reproduce with it. That has been freeware, if you like. That's been the understanding," said Suleyman. However, several ongoing copyright lawsuits and many artists and content creators would disagree with Suleyman’s remarks.

Goldman Sachs says the return on investment for AI might be disappointing
According to a new report from Goldman Sachs, tech companies plan on spending over $1 trillion on AI, most of which will go into building new data centres, the power grid, and AI chips. However, the report warns that the return on investment may take a long time and could be disappointing, stating that “this spending has little to show for it so far.” Other experts quoted by Goldman Sachs were more enthusiastic, suggesting that the technology will get cheaper and will follow a similar path to that of smartphones, Uber, or Airbnb.

Zuckerberg disses closed-source AI competitors as trying to ‘create God’

Mark Zuckerberg recently gave an interview in which he outlined Meta’s AI strategy and how Meta’s AI products could assist creators with tasks like using AI agents to represent the creator or small businesses. Zuckerberg shared his vision for an “AI-powered future” for artists and creators, as well as the future of Meta, which includes smart glasses and neural wristbands to replace smartphones. He also addressed his closed-source competitors, stating that AI should not be a technology hoarded and controlled by one company. “It’s almost as if they kind of think they’re creating God or something and … it’s just — that’s not what we’re doing,” he said. “I don’t think that’s how this plays out.” Additionally, Zuckerberg revealed that Meta is working on Llama 4 which looks like will be multimodal and be better at reasoning.

---

If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?

---

🤖 Robotics

▶️ Figure Status Update - BMW Full Use Case (1:38)

In January this year, Figure announced a partnership with BMW, in which the German car manufacturer will test Figure’s humanoid robot in one of its factories in South Carolina. In this video, Figure provides an update on that partnership, demonstrating how effective their robot is at manipulating objects—sometimes two at a time—and placing them precisely where they need to be.

Persona AI Brings Calm Experience to the Hectic Humanoid Industry
A new player has entered the humanoid robotics game. Persona AI is a company founded by Nic Radford, who led the team that developed NASA’s Valkyrie humanoid robot, and Jerry Pratt, former CTO of Figure, one of the biggest players in the humanoid robotics scene. In this interview with IEEE Spectrum, they share why they founded yet another humanoid robotics company, their plans for Persona AI, and how they intend to tackle significant challenges such as robotic hands, batteries, and safety.

Japan introduces enormous humanoid robot to maintain train lines
West Japan Railway has introduced a new humanoid robot for maintenance work on the company’s network and to address worker shortages. This 12-metre-high robot, with a small Wall-E-like head, is mounted on a truck that can drive on rails while its operator, who sits in a cockpit on the truck, "sees" through the robot’s eyes via cameras and remotely operates the robot’s powerful limbs and hands. The machine can use various attachments for its arms to carry objects as heavy as 40 kg (88 lbs), hold a brush to paint, or even use a chainsaw.

Robotic hand with tactile fingertips achieves new dexterity feat
A team at the University of Bristol has developed a four-fingered robotic hand with tactile fingertips capable of rotating objects in any direction, even upside down. Unlike other robotic hands, this new hand uses simpler, cost-effective methods and desktop computers. The breakthrough comes from integrating high-resolution tactile sensors into the robot hands, inspired by advances in smartphone cameras. Bristol's artificial fingertips use a 3D-printed mesh that mimics the structure of human skin. The next steps include advancing to tasks like assembling items, moving beyond basic rotation and pick-and-place operations.

Why Not Give Robots Foot-Eyes?
Meet Foot Vision, a new sensor developed by researchers at Southern University of Science and Technology in Shenzhen. Attached to the feet of the Unitree Go1 robot dog, Foot Vision is essentially a camera that helps the robot better sense the surface it walks on. Preliminary results showed that Foot Vision enabled the robot to perceive the flow of sand or soil around its foot as it takes a step, which can be used to estimate slippage. This valuable information can help the robot navigate various terrains more effectively.

🧬 Biotechnology

Biotech Behemoths in Review
Here is a very detailed report taking a closer look at 50 biotech startups (mostly US-based) founded in the last 15 years that have recently generated tremendous value for patients, investors, and the sector. The report examines their product impact, the stages of clinical trials their products are in, and each company’s market valuation. It also takes a look at the characteristics of their founders and the origins of the companies’ key technologies or drugs. Additionally, the report compares these “Biotech Behemoths” to top tech startups from the same time period. Overall, the report provides valuable insights into the mostly US biotech startup scene.

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!"