Cannibals, Commodities, and Conquest

DeepSeek and the Grand Strategy to Capture the Commanding Heights of AI

Audere Capital invests in exceptional founders building advanced technology solutions to solve critical problems. FOLLOW AUDERE ON LINKEDIN

January Recommended Reading:

• Creative Humans: How Technology is Transforming Human Nature and Possible Futures

• What Tech Calls Thinking: An Intellectual History of Silicon Valley

• Chip War: The Fight for the World’s Most Critical Technology

"For there is no folly of the humankind that has not, in its own moment, seemed the height of wisdom"
- Moby Dick, Herman Melville

"Now is no time to think of what you do not have. Think of what you can do with what there is."
- Old Man and the Sea, Ernest Hemingway

"You’re going to need a bigger boat…"
- Chief Brody, Jaws

DeepSeek’s emergence in the artificial intelligence market has been nothing short of revolutionary. At its core, DeepSeek’s model represents a stark departure from conventional AI development strategies, both technologically and economically. But the significance of their new R1 model is far greater than simply its capabilities. It represents a radical new strategic dimension in the AI market.

A Revolution From the Deep in AI Technology

From a technology perspective, what sets DeepSeek apart is its ability to deliver exceptional performance while eschewing reliance on proprietary, high-cost hardware. Instead, it operates seamlessly on commodity chips—the kind commonly found in consumer-grade devices or mid-tier servers.

Most AI systems, particularly high-performing models like OpenAI’s GPT-4, rely heavily on specialized GPUs from Nvidia, whose market dominance is both a strength and a vulnerability. These GPUs are not only expensive but also come with supply chain dependencies that can stymie scalability. DeepSeek sidesteps these limitations entirely, leveraging innovative algorithmic techniques that optimize resource allocation, memory management, and computational efficiency. The result is a model that rivals—and in some cases outperforms—its peers, at a fraction of the cost.

Additionally, DeepSeek has embraced a bold open-source ethos. And this is actually a very critical decision. By releasing its codebase, DeepSeek has democratized access to cutting-edge AI capabilities, ensuring that small businesses, researchers, and even hobbyists can deploy advanced AI without prohibitive costs. This approach has catalyzed widespread adoption (it’s already the number one app on the Apple Store) while simultaneously destabilizing the economic foundations of traditional AI market leaders.

In short, DeepSeek has undermined, potentially permanently, the investment approach and the hundred-billion-dollar valuations of many of the leading AI competitors, not to mention Nvidia’s share price. The decision to open-source its platform is as strategic as it is disruptive. By accelerating adoption and encouraging integration into diverse workflows, DeepSeek has positioned itself as an essential building block in the AI ecosystem.

IRRELEVANT SAFETY BLANKETS

It is a hallmark of how threatened they feel that many of the bigger players in the market have been behaving as if they saw this coming, and that it is in fact a good thing. To be clear, it was predictable, and it is a good thing. But not for them. Definitely not for them.

In response to DeepSeek’s disruptive entry, many industry leaders, including Microsoft’s Satya Nadella, have pointed to Jevon’s Paradox as a way to rationalize their optimism. Rooted in the 19th-century economic observations that while increases in coal mining efficiency drove commodity prices down, these declines were more than overcome by the resulting increased consumption.

Jevon’s Paradox is ultimately just a fancier way of saying that efficiency lowers prices but the lost revenue can be regained, and increased, by growing demand. Unfortunately, in this case time is required to understand how this will manifest in use and in markets, rather than just repeating what we learned in econ 101.

Applied here, the logic is that as DeepSeek lowers the cost of AI, adoption will surge, and the total market size will grow exponentially. And that is obviously true. When gas is cheap, people drive more and oil companies make more. But this observation is irrelevant to the current moment.

Javen’s Paradox fails to address the more pressing issue: who will capture the lion’s share of this expanding AI market? What will prove to be the long-term valuable AI capability? Chips, Algorithms, Data, Distribution, Edge Devices…?

Javen’s Paradox fails to address the only thing that matters: What will prove to be the long-term valuable AI capability? Chips, Algorithms, Data, Distribution, Edge Devices…?

Last week, we thought it was OpenAI. There was a press conference with Softbank (not necessarily a strong leading indicator but more on that another day) about deploying an additional $500B into the space. But history is rife with examples of industries where innovation expanded the market but obliterated profitability for incumbents.

For a recent (and we’re dating ourselves) instance, take the rise of personal computers in the 1990s. It was clear by the mid-1980’s that PCs were going to have a foundational role in the global economy of the future. But as it played out, the market was far more complex and only a handful survived. Understanding how they did is key to anticipating the AI winners.

VALUE CREATION VS. VALUE CAPTURE

In our mind, too much attention is focused by entrepreneurs and investors on the consulting sounding ideas of “creating value” and not enough on the capital returning necessity of “capturing value.”

The outcome of the PC wars was not an accident. In new markets, the evolution is not always random, sometimes it is planned.

The early days of the personal computer era offer a cautionary tale. In the 1980s, giants like IBM, HP, and Compaq were well-positioned to capitalize on the booming demand for PCs. The market appeared limitless, and growth was strong across the board. Yet by the end of the decade, it became clear that most hardware manufacturers were locked in a zero-sum game, their margins eroded by intense competition.

The outcome of the PC wars was not an accident. In new markets, the evolution is not always random, sometimes it is planned.

STEP 1: THE RISE OF THE CANNIBALS

Throughout the 1980s and 1990s, a peculiar reality emerged in the personal computer market: if you bought a PC, it was almost guaranteed to run on Microsoft’s Windows operating system and be powered by an Intel chipset. This dominance was remarkable because neither Windows nor Intel processors were visible to the consumer. These were intangible components, buried deep in the device, yet they became the most recognizable and trusted brands in the PC ecosystem.

This dominance did not arise by chance. Microsoft and Intel carefully shaped the economics of the PC market. They recognized that the average consumer was willing to spend around $1,500 on a personal computer. This became the de facto price point for the industry, providing a cap on the total value that could be extracted from each sale. However, this also meant that competition within the ecosystem would determine how much profit any participant could claim.

Realizing this, Microsoft and Intel devised a strategy to maximize their share of this fixed profit pool. They became Cannibals. Not of their own companies, but of their industries.

The result was the infamous Wintel strategy: a deliberate partnership between Microsoft, the dominant software provider, and Intel, the leader in chipsets. Together, they established their technologies as industry standards, ensuring that every PC was effectively locked into running Windows on Intel processors.

STEP 2: COMMODITIZE YOUR COMPLEMENTS

The Wintel strategy relied on commoditizing every other part of the value chain. Microsoft and Intel drove industry-wide standards that made it easy for multiple hardware manufacturers to produce interchangeable components—motherboards, hard drives, RAM, and peripherals all became commodities. By setting benchmarks that were widely achievable, they ensured intense competition among manufacturers, driving down prices and margins for every player except themselves.

Simultaneously, Microsoft and Intel invested heavily in branding and marketing to make their offerings synonymous with reliability and performance. Microsoft’s Windows became the default choice for operating systems, while Intel’s “Intel Inside” campaign elevated the processor to a recognizable and desirable consumer feature. This branding effort solidified their position as indispensable players in the PC ecosystem.

STEP 3: SCALE AND SPEED TO CONQUER

The Wintel strategy was devastatingly effective. By commoditizing hardware components, Microsoft and Intel forced the rest of the ecosystem to operate on razor-thin margins. PC manufacturers like Dell, Compaq, and HP competed fiercely on price, while Microsoft and Intel captured disproportionate value as the gatekeepers of software and chipsets.

The strategy also established a virtuous cycle. As more PCs were sold with Windows and Intel processors, developers standardized their software to run on this platform, reinforcing the dominance of the Wintel ecosystem. Over time, Microsoft and Intel’s grip on the market became unassailable, allowing them to dictate the terms of the industry and collect outsized profits while their partners scrambled to survive on increasingly narrow margins.

The lesson from Wintel is clear: by commoditizing complements and controlling the most critical layers of the value chain, Microsoft and Intel achieved conquest in the PC market. This same dynamic is now playing out in the AI market, where the question of who will capture value hinges on similar strategic considerations.

WHAT HATH DEEPSEEK WROUGHT?

Microsoft and Intel’s Wintel strategy provides a blueprint for understanding DeepSeek’s approach. The duo recognized that the total value of the PC market was constrained by a relatively fixed price point of around $1,500. To maximize their share of that value, they drove commoditization in every layer of the value chain except their own. By ensuring that hardware components were interchangeable and software compatibility was ubiquitous, they created an ecosystem where no one else could command significant margins.

DeepSeek appears to be employing a similar playbook. By commoditizing AI hardware requirements and making its software freely available, it has dramatically lowered the barriers to entry for deploying AI solutions. This strategy disrupts competitors like OpenAI, whose reliance on expensive infrastructure makes them vulnerable to cost-based competition. At the same time, it positions DeepSeek as a foundational layer in the AI stack, potentially enabling it to capture downstream value through services, integrations, and premium features.

THE THREE WAVES OF AI INNOVATION

DeepSeek’s strategy can also be viewed through the lens of the "Three Waves of Innovation" framework, which we’ve discussed in our prior post “Fourteen Elephants on a Steel Bridge.” The first wave, or the "Extract Wave," is driven by performance gains and cost efficiencies. DeepSeek’s ability to deliver cutting-edge performance on commodity hardware exemplifies this phase. By extracting value from legacy AI providers tied to expensive infrastructure, DeepSeek is reshaping the economic landscape of the industry.

The second wave, or the "Expand Wave," occurs when new technologies unlock previously inaccessible markets. DeepSeek’s open-source approach has the potential to catalyze this wave, enabling small businesses, emerging markets, and nontraditional users to adopt AI at scale. This democratization could expand the total addressable market for AI solutions by an order of magnitude.

The third wave, or the "Entrench Wave," is where the real battle for value capture begins. In this phase, technologies become standardized and embedded into broader ecosystems. The players who dominate this wave are those who position themselves as indispensable platforms. If DeepSeek can transition from being a tool to becoming the foundation of countless applications, it will have successfully captured the strategic high ground.

CONQUEST AMID COMMODITIES

History teaches us that commodification is both a threat and an opportunity. For incumbents, it often leads to margin compression and strategic paralysis. For challengers like DeepSeek, it opens pathways to redefine value chains and establish new paradigms.

In nascent markets like AI, value will migrate over time. The first place it is created is more likely than anything to not be where the ultimate prize lies. And capturing value is not merely a function of technological moats but of the strategic tactics deployed by market participants. Companies are actively fighting on multiple fronts and at multiple layers—hardware, algorithms, applications, and ecosystems—to define, shape, and capture as much value as possible.

DeepSeek’s decision to commoditize certain layers of the AI stack while positioning itself as indispensable elsewhere reflects an understanding of these dynamics. The critical question is whether it can sustain this balance and transition from a disruptor to a dominant platform.

RED IN THE MORING, AI’S WARNING

DeepSeek’s impact on the AI industry is a vivid reminder that the rules of technological markets are never static. The company’s bold gambit to commoditize AI has upended conventional wisdom and forced competitors to rethink their strategies. Whether it emerges as a Microsoft-like giant or a cautionary tale will depend on its ability to balance value creation with value capture.

In the meantime, one thing is clear: the AI landscape is undergoing a seismic shift. For investors, entrepreneurs, and policymakers, the lessons of history offer invaluable guidance. In the words of Marshall McLuhan, "We march backward into the future."

The past is not just a prologue; it is a map for navigating the uncharted terrain of technological revolutions. And as we have noted before in our post on Interregnum periods of rapid change: the strong do not defeat the weak; the fast defeat the slow.

In periods of rapid change, the strong do not defeat the weak; the fast defeat the slow.

SEEING PAST THE HORIZON

There remain more questions than answers at this point and they are important ones. Among others:

• Will this open source innovation actually create a widely embedded vulnerability if, as it seems likely, it becomes rapidly adopted across tech stacks? At a minimum it has already been demonstrated to have strong embedded editorial control. We are already seeing evidence of selective edits, as in Tiananmen Square. What else is not included, or what else in embedded?

• What kind of digital visibility does this give to a nation that has a complicated history with aggressive thought policing, restrictive free speech, international advocacy in contested regions, and a tense relationship with much of the rest of the world?

• What does it mean for many of the technocrats so recently in ascendency that they’ve had their technological empires so deeply and unexpectedly undermined? Even if new technological advances are achieved, the economics of their empires have been significantly and possibly permanently impaired.

As always, build boldly and dare to be first.

Peter & Maggie