Unlocking AI's True Potential: DeepSeek's Open Source Breakthrough
The Illusion of AI Democratization: Big Tech's Controlled Open Source
Chips can be restricted, open source cannot.
Tech giants have long championed "AI democratization," but in practice, they've only pursued democratization that doesn't threaten their core business interests. Although major tech companies have various open-source initiatives, they deliberately release AI models that are 6-12 months behind their latest versions to prevent competitors from replicating their core technologies and threatening their business models.
Thus, despite years of rhetoric, true AI democratization hasn't materialized: On the hardware side, chips remain prohibitively expensive and scarce for average users; on the software side, the latest technologies remain proprietary due to competitive dynamics among Big Tech companies, as none want to aid competitors or compromise their own position.
Meta's Llama: A Strategic Approach to Open Source AI
Meta's Llama appears to be the sole exception. Meta's seemingly noble approach to openness stems from their assessment that open-sourcing Llama won't harm their core business interests. Instead, it positions Meta as an open-source AI leader, enabling them to expand their ecosystem further. (Meta's core competitive advantage lies not in any particular technology but in their users' inability to escape the gravitational pull of network effects – a topic for another discussion.)
The AI Supply Chain Monopoly: A Tight Grip by Tech Giants
Consequently, the AI supply chain remains firmly controlled by a handful of tech companies:
Nvidia → ASML → TSMC → Big Tech → Enterprises and Consumers
Their market shares in respective domains are:
* Nvidia (chip design): 85-90%
* ASML (semicon manufacturing equipment): 100%
* TSMC (foundry services): 90%
* Big Tech: 70%
Big Tech primarily provides AI computing power and applications through cloud services, with U.S. cloud providers commanding over 70% of the global market share.
These near-monopolistic market shares indicate the chain's robustness. As gatekeepers of enterprise and consumer AI needs, Big Tech drives chip demand. This prompted Sequoia Capital's notable June article, "AI's $600B Question," warning that while massive investments flow into AI, demand for AI chips primarily comes from Big Tech's purchases from Nvidia, raising questions about actual consumer applications and market sustainability.
Big Tech's Strategy: Chip Dominance and AI Leadership
Two years after ChatGPT's debut, upstream hardware vendors like Nvidia have secured profits, while Big Tech software companies are simultaneously investing heavily in AI infrastructure and gradually integrating AI into their existing business models as value-added services: "We're not rushing to find gold, but we're buying all the shovels; this way, even if others strike gold, they'll do so using our rented tools."
The formula for U.S. AI industry dominance has been: "Cornering the market on advanced AI chips + closed-source leading AI software."
DeepSeek's Disruption: Low-Cost Deployment and Open Source
DeepSeek's "low-cost deployment + open source" approach directly challenges both elements of this formula.
This has destabilized the AI supply chain, as open source enables consumer demand to originate from end-users rather than being controlled by tech giants. Simply put, DeepSeek R1 can now be deployed on a mid-range graphics card, eliminating dependence on expensive cloud-based AI models from tech giants. This alone could trigger massive demand for edge AI applications, accelerating AI adoption across sectors.
The Power of Open Source: DeepSeek's Ecosystem and Global Impact
Open source is often associated with democratic values like "freedom and openness" and viewed as "noble, beneficial, selfless, and altruistic." Thus, when a non-democratic nation produces history's most significant AI open-source project, it naturally creates discomfort and contradiction in the Western world.
However, open source remains open source. Despite skepticism toward DeepSeek, anyone can modify it to their preferences or build upon it to develop even more powerful models. Regardless of U.S. responses, developers worldwide outside China and the U.S. have begun building new ecosystems around DeepSeek R1 (developer activity on Hugging Face and Github has been remarkably high this past week) and continuing to optimize and open-source further.
Open Source AI: The Strategic Crossroads for Tech's Future
Open source cannot be contained. The U.S. should focus less on restricting Nvidia's chip sales and more on pressing Big Tech to clarify their future open-source strategies.
This presents a challenging situation because once something is open-sourced, it's irreversible. The battlefield has shifted – AI demand now stems from open-source software and global developers rather than chip supply. It's crucial not to misidentify the key players in this new landscape.
(This article is translated from a Facebook post by Sega Cheng, co-founder and CEO of iKala.)
