Contrary Research Rundown #96
Sacrificial AI offerings to the hyperscaler Gods, plus new memos on Anthropic, Kajabi, and more
Research Rundown
Technological paradigm shifts have, in the past, created an opening for new entrants to dominate. The introduction of SaaS in the late 90s and early 2000s enabled Salesforce’s CRM to race ahead of incumbent on-premise solutions like SAP and Oracle. The rise of the internet enabled digital platforms like Amazon and Netflix to surpass brick-and-mortar incumbents like Borders and Blockbuster. The introduction of cloud computing enabled Amazon and Microsoft to build dramatically larger computing businesses than IBM or Oracle.
The current AI boom is lauded as the latest paradigm shift. And technologically, that seems obvious. Some estimates indicate that generative AI has the potential to add $4.4 trillion in annual value, with half of all work activities getting automated by 2045. But at the same time there is a much more limited “leap frogging” going on between incumbents and new entrants.
The immediate counterpoint to that is OpenAI. The company was estimated to be generating $3.4 billion in revenue as of June 2024. And there are certainly some extenuating circumstances with OpenAI. The company benefits from a sort of “status quo” effect — nobody gets fired for buying OpenAI’s products.
At the same time there is a clear and obvious feeling of OpenAI being beholden to Microsoft. And that’s not even reading between the lines. If you remember one surprising piece of posturing from Satya Nadella, the CEO of Microsoft, it was how explicitly he made OpenAI sound dominated:
"If OpenAl disappeared tomorrow, we have all the IP rights and all the capability. We have the people, we have the compute, we have the data, we have everything. We are below them, above them, around them."
Unlike any prior paradigm shift before the incumbents who are intimately connected to AI continue to demand sacrifices of the companies striving to innovate in the space. At first, it was Microsoft paying $650 million in March 2024 for a “licensing” deal to use Inflection’s models. Oh, and by the way, the deal let Microsoft hire away most of the company’s staff and co-founders. Next, it was Amazon inking an identical deal with the team at Adept (though its unclear if they even had to pay anything.) Most recently it was Google’s deal with Character AI to buyout the company’s investors at 2.5x the company’s $1 billion valuation from 2023.
Predictions for what’s next? Mistral acquired by Meta’s AI team?
Several of these deals were poised as “license and hire” vs. acquisitions in order to avoid regulatory scrutiny. Despite the attempted maneuver, the FTC is still taking note. Thus far the Microsoft and Inflection deal has come under investigation, but it could very easily extend to the others as well.
The combination of massive costs for compute, training data, and talent seem to be sizable hurdles for anyone other than the incumbents, or those who’ve been bequeathed their power, like OpenAI, to succeed. Companies like Glean, Cohere, Writer, Harvey, and Hugging Face are tripling or quadrupling their revenue. But these have anywhere from $30 to $100 million in revenue; a far cry from the billions the incumbents are generating.
And what’s more, firms like Sequoia are cautioning the potential bubble that exists in AI, pointing the need for $600 billion of AI revenue to materialize to justify the investment in the space. Massive hedge funds, like $70 billion Elliot Management, are saying Nvidia is “a bubble” and that AI is “overhyped.” Companies like Microsoft and Meta are hedging their bets, saying it could take 15 years for these investments to pay off. In fact, Microsoft is already missing revenue targets around its AI business. And the investment certainly isn’t slowing. Alphabet, Microsoft, and Meta are heavily investing in AI, with a combined expenditure of $40.5 billion on infrastructure, land, and chips in Q2 2024 alone.
The question now will be whether there are still meaningful opportunities to build sizable AI businesses independent of the existing incumbents? Or if this will just be an extension of the cloud wars playing out in public? When you have Google employees leaving to start an AI company and then getting “acqui-hired” back three years later… it feels like it might be the latter.
Kajabi provides tools for creators to directly monetize their audience through digital content and services. To learn more, read our full memo here and check out some open roles below:
Senior Software Engineer, Monetization - Remote (Brazil)
Data Engineer Lead - Remote (US)
Anthropic is an AI safety and research company working to build reliable, interpretable, and steerable AI systems. To learn more, read our full memo here and check out some open roles below:
Software Engineer (API Experience) - Remote-Friendly (Travel-Required) or San Francisco, CA or New York City, NY
Software Engineer (Infrastructure) - London, UK
Check out some standout roles from this week.
Resourcely | San Jose & Santa Clara, CA - Frontend Engineer, Software Engineer
Baseten | San Francisco, CA, New York, NY or Remote (US) - Software Engineer -Infrastructure, Software Engineer - Fullstack, Forward Deployed ML Engineer
Railway | Remote (anywhere) - Senior Infrastructure Engineer, Senior Full-Stack Engineer - Product
Carta's COO is departing the company just one year after joining.
OpenAI is attempting to be more than a premium model offering. Just a year ago, GPT-4 cost $37.5/1M tokens. Now, GPT-4o mini costs $0.3/1M tokens.
Former Cisco CEO John Chambers is joining the board of Humane, which is known for its wearable device that aims to replace smartphones.
Open-source AI models are rapidly closing the performance gap with proprietary models, while offering significant cost advantages. Open-source AI models like LLaMa 3.1 are now competitive with leading proprietary models like GPT-4, while being over 7x cheaper.
Canva announced its acquiring Leonardo.Ai, a fast-growing AI startup with over 19 million users and its own foundation AI model which launched in December 2022. The deal bolsters Canva's AI capability ahead of an anticipated IPO in 2025.
The nuclear industry is trying to reinvent itself with smaller, modular, and advanced reactor designs to meet growing energy demands.
Cyber firm KnowBe4 hired a North Korean threat actor posing as a U.S. software engineer using stolen credentials and an AI-generated photo.
Perplexity AI announced a partnership program with several media companies that aims to share advertising revenue.
Wiz, a cloud security startup, turned down Google's $23 billion acquisition offer, believing it can grow even bigger and reach $1 billion in ARR before an IPO.
Microsoft's Azure AI and GitHub Copilot are seeing strong growth, with Azure AI at $5B ARR growing 900% y/y and Copilot at $300M ARR growing 180% y/y.
Vendr published it’s Q2 SaaS Trends Report, where they present data-backed insights to save time and money on software.
Alex Kolicich, founding partner at 8VC, published Q2 2024 State of Venture Update. In it he explains that AI valuations are fully priced, but the value capture remains uncertain, cloud software companies are undervalued relative to their potential to benefit from AI, there may be an oversupply of GPUs, which could signal a "trough of disillusionment" for AI in the near future and government reforms have made it easier for defense tech startups to compete and succeed in the defense industry.
In this month’s 10 Charts That Capture How the World Is Changing, Rex Woodbury covers a broad range of topics, from ketamine to WhatsApp users, egg freezing to AI job loss.
Open source startup FOSSA has acquired developer community platform StackShare, which has over 1.5 million registered developers, in a deal that aims to help FOSSA expose its governance and security tools to a broader set of developer tools.
On Wednesday July 31, Nvidia closed 13% higher, adding $329 billion to its market value in a single day. The largest single day jump in market value ever.
Industrial robotics software maker Covariant has received takeover interest from Amazon.
Elon Musk's AI startup xAI has reportedly considered acquiring chatbot maker Character.AI as it looks to expand its AI capabilities and test its Grok AI models, though Musk denies the reports.
Applied Materials was denied a $4 billion Chips Act grant for a research center in Silicon Valley, dealing a critical blow to the company's plans.
During Meta’s Q2 earnings call, CEO Mark Zuckerberg, said the latest Llama 3 model is competitive with the most advanced AI models and leading in some areas, with expectations that future Llama models will become the most advanced in the industry. “I think the team there is doing really great work going from the first version of Llama, to Llama 2 last year that was a generation behind the frontier and now Llama 3.1, which is basically competitive and, in some ways, leading the other top closed models.”
During that same earnings call, Zuckerberg said "The amount of compute needed to train Llama 4 will likely be almost 10x more than what we used to train Llama 3 -- and future models will continue to grow beyond that.”
Intel is reportedly cutting up to 19,650 employees and suspending its dividend.
Balaji tweeted, “tech guys don't use Google search anymore. It's not just censored. It just sucks. Perplexity is better.”
Avi Schiffmann, the former high school student who created a popular COVID-19 tracking website, has now raised $2.5 million to launch “Friend” — an AI-powered wearable device meant to be a digital companion that users can talk to and develop a relationship with.
At Contrary Research, our vision is to become the best starting place to understand private tech companies. We can't do it alone, nor would we want to. We focus on bringing together a variety of different perspectives.
That's why we're opening applications for our Research Fellowship. In the past, we've worked with software engineers, product managers, investors, and more. If you're interested in researching and writing about tech companies, apply here!
Great roundup this week guys