A Physical Social Network℠

Frank Albanese and Ben Schwartz

Original version published on February 2, 2025.

In a seminal scene in Iron Man, the antagonist grows frustrated at the inability of his engineering team to replicate Tony Stark’s miniature arc reactor. He berates his head engineer, “Tony Stark was able to build this in a cave! With a box of scraps!”

Chinese startup DeepSeek pulled a Tony Stark-like engineering marvel with the release of its newest AI model, leaving U.S. AI companies and investors dumbfounded, just like the engineers in Iron Man. DeepSeek’s model was built faster than its Western counterparts, on less-advanced microchips, using 10% of the computing power — at a fraction of the cost. It’s a sobering moment for the U.S. tech sector. Even Marc Andreessen, general partner of Andreessen Horowitz, one of the biggest U.S. firms investing in AI, labeled the breakthrough a “Sputnik moment,” analogous to the event that spurred the space race.

And, as in 1957, the news has already had a major impact on public confidence in the future of American technology. Following DeepSeek’s announcement, major AI-heavy stocks plummeted. AI chip-maker Nvidia lost $593 billion in market value — the largest single-day loss for one company in Wall Street history. Microsoft, Google, and Broadcom also saw significant declines, with semiconductor indexes like SOX dropping 9.2%.

The current AI-driven boom bears striking similarities to past speculative bubbles. The dot-com era saw vast sums invested in internet startups that struggled to deliver sustainable returns. More recently, the cryptocurrency craze captivated investors, only to falter as the technology failed to justify its valuations.

Investors have poured billions into Sam Altman’s OpenAI, just to find more advanced technology is now open source for all to use, courtesy of DeepSeek. This breakthrough serves as a dramatic warning: watershed advancements in AI can manifest and change the status quo of the entire industry overnight.

These stock market shocks reflect a deeper anxiety among investors: are the returns from AI truly capable of justifying the enormous capital inflows? As the hype cycle around AI reaches an inflection point, the industry is at a crossroads. Overfunded AI companies now face a stark choice: innovate rapidly to meet sky-high expectations or collapse under the weight of their valuations.

Another related sector of business that has been heavily invested in may also be transformed: B2B SaaS.

Venture capitalists have invested heavily in B2B SaaS over the last fifteen years. The model is irresistible: sell software as a subscription service to enterprises with deep pockets, scale the product globally, and enjoy recurring revenue streams. While coding is a human endeavor, building software requires significant time, money, and expertise, making it worth outsourcing. In 2023, forty-seven percent of U.S. venture capital investments flowed into SaaS, with 88% of that focused on B2B solutions. In 2024, SaaS spending averaged $8,700 per employee annually.

But DeepSeek’s breakthrough has demonstrated that the technological barriers to creating sophisticated AI tools are shrinking — fast. Within five to ten years, AI tools may be able to generate enterprise-class platforms by themselves.

The B2B SaaS industry is based on the premise that it is cheaper, faster, and better to outsource a business’s software needs to a company specifically dedicated to creating the platforms that allow a business to run — CRMs, HR, billing, etc.

Once companies have AI with the capability to mimic existing B2B platform functionalities at scale, an “insourcing” model will be very appealing: having their own AI develop software will not only be cheaper than outsourcing it to SaaS companies, but will also allow businesses to customize functionality to fit their particular use cases, posing an existential threat to the SaaS business model as we know it.

SaaS companies are going to have to evolve. As their technology becomes easier to replicate, they will need to emphasize other sources of added value, including superior service, network effects, and proprietary data. Pricing will also need to adapt to a world in which outsourced software complements in-house enterprise software, rather than dominating the market.

For B2B SaaS investors, however, the implications of DeepSeek’s announcement could be even greater. DeepSeek’s breakthrough demonstrates that the rapid advancement of this type of technology can occur even without major investment. As these advances begin to squeeze B2B SaaS profit margins, investors may double down on their bets, expecting even more innovation from entrepreneurs to make up the now clearly inflated value proposition.

Or perhaps they will take DeepSeek’s release as a signal to start to look elsewhere for ROI. Investors may consider diversifying — finding an industry vertical that is ripe for disruption or supporting new business models — perhaps even returning to B2C, which has suffered a dip in investment in recent years.

When everyone can build advanced software out of a box of scraps, it’s not just AI giants that will have a rude awakening. Tectonic changes could be taking place in entire industries and customer segments that will change how investors allocate capital for years to come.