In the past decade, a striking alliance has formed between Silicon Valley and Wall Street, culminating in a classic narrative of a dual oligopolySilicon Valley drives the story through technological breakthroughs, while Wall Street amplifies expectations through capital leverageHowever, as this narrative unfolds, it requires constant refreshment, necessitating the introduction of new materials and plotlinesOver the last two years, artificial intelligence (AI) has emerged as the prominent protagonist in this unfolding drama.

However, a twist in this narrative is anticipated by 2025.

While tech giants such as Google and Meta continue to commit billions of dollars to building colossal computing power clusters, a new player called DeepSeek has demonstrated that equivalent performance can be achieved with just one-third of the computational resourcesThis reality underscores a shifting paradigm: an accumulation of computing power is no longer the singular pathway to intelligent evolution; instead, algorithmic efficiency is evolving into a new valuation metric.

The rise of OpenAI exemplifies this new model.

Founded in December 2015, OpenAI was valued at just $1 billion by 2019. However, the launch of ChatGPT in late 2022 catapulted its valuation to around $20 billion

Advertisements

Following an impressive $6.6 billion financing round in October 2024, the company now boasts a staggering $157 billion valuation—yet profitability remains elusive.

Other leading AI companies are also leveraging high valuations to secure massive fundingFor instance, xAI, founded by Elon Musk in 2023, raised $6 billion in its C funding round in December 2024, achieving a valuation exceeding $40 billionAnthropic is seeking $2 billion in a new funding round with a valuation of $60 billion.

This model operates on the fundamental logic that technological breakthroughs require vast funding to support computing power and talent investments, while capital market valuations hinge on the perceived “moat” of technology.

According to a report released by PitchBook, an American venture capital data research company, AI startups in the U.Ssecured approximately $97 billion in venture capital funding in 2024, nearly half of the total $209 billion raised across all startups—most of which hail from Silicon Valley.

The open-handedness of venture capital firms is rooted in a well-founded belief that AI companies emanating from Silicon Valley are backed by a robust capital market poised for substantial returns, and indeed, this belief is proving correct.

Even after a recent downturn over the past fortnight, companies like Tesla and Nvidia continue to maintain high valuations

Advertisements

Currently, Tesla's price-to-earnings ratio stands at a staggering 170 times; investors perceive it less as a mere electric vehicle manufacturer and more as an AI entity, as reiterated by MuskNvidia's price-to-earnings ratio is at 45 times, while Microsoft sits at 33 times.

The seven giants of the American stock market require fresh narratives and new plot developments to keep their investors enthralled with compelling storiesFor the past two years, the linchpin sustaining their lofty valuations has been AIAs such, the “technology-capital” narrative has transformed into an “AI-capital” framework.

However, this “AI-capital” narrative carries inherent fragilities, as its sustainability is entirely reliant on a positive feedback loop between “technological leadership” and “capital returns.” The moment DeepSeek achieved comparable model performance to OpenAI's hundreds of millions of dollars worth of investment at a cost of just $5.5 million, the mythical narrative of AI concocted by Silicon Valley and Wall Street faced widespread skepticism.

More critically, DeepSeek's open-source strategy directly undermined the scarcity premium associated with closed-source models, a key variable in the “AI-capital” valuation model.

The paradigm shift brought about by low-cost AI is poised to alter the landscape dramatically.

DeepSeek's disruptive innovation stems not from a generational leap in technology but rather from the supreme optimization of existing resources

Advertisements

Its V3 model enhances the output per unit of computational power by more than twofold utilizing heterogeneous computing architectures and dynamic load balancing algorithmsAs a result, training costs are just a fraction of the industry average, paving the way for a new, low-cost, high-efficiency technological pathway.

The “efficiency revolution” initiated by DeepSeek has laid bare the shortcomings of Silicon Valley’s approach, which has traditionally favored a strategy of merely piling on computing power and dataFor example, Musk has indicated that the training costs for OpenAI's in-development GPT-5 model have already exceeded $1 billion, while DeepSeek-R1 achieved equivalent performance at a cost of only $600,000—significantly lowering the industry valuation benchmarks.

This efficiency revolution has triggered a ripple effect throughout the capital markets

• India Cuts Rates by 25 Basis Points
• How to Navigate the AI Investment Boom?
• Amazon Doubles Revenue and Profit in Q4
• Global Public Debt on the Rise
• The Illusion of Ever-Rising UK Stock Market

Nvidia's stock price plummeted by over 20% within a week, largely reflecting a temporary flight to safety amid an overvaluation crisis, but it also indicates a growing distrust in the “burn cash for growth” paradigmThis marks a crack in the AI narrative model condoned by Silicon Valley and Wall Street.

The story of “AI-capital” is essentially a collusion between “technological idealism” and “capital expansion needs.” Microsoft’s plan to invest roughly $80 billion in AI data center construction in the 2025 fiscal year, alongside OpenAI's collaboration with Nvidia and SoftBank on the $500 billion “Interstellar Gateway,” underscores this collusion.

DeepSeek’s historic contribution lies in its commitment to open-source methodology; it has practically demonstrated the feasibility of high-cost-performance AI products

This move enables a surge of small and medium-sized enterprises and developers to join the technological fray, shifting the AI application ecosystem from an “oligopoly game” to an inclusive “everyone participates” model.

Consequently, the grand AI narrative spun by Silicon Valley and Wall Street is losing some of its former luster.

As the dust settles, it is evident that Silicon Valley and Wall Street are struggling to cling to their original narrative logic.

For instance, OpenAI and Meta are attempting to maintain their technical authority through “debunking” tactics, but they have met with ridicule; Anthropic executives have called upon U.Sauthorities to tighten chip export regulations, seeking to leverage geopolitical maneuvers to uphold the technical gap.

Nevertheless, such resistance is unlikely to alter the fundamental trend

The DeepSeek phenomenon exposes not only a new choice of technological paths but also signifies a paradigm shift in industry logic: open-source is swiftly becoming an unstoppable wave in AI modelingThe decentralized ecosystem it promotes threatens to incessantly assault the closed-source dominance traditionally held by Silicon ValleyEven OpenAI’s CEO Sam Altman has conceded that the company’s closed-source strategy stands “on the wrong side of history.”

Recent statements by IBM's CEO Arvind Krishna highlighted a commonly held misconception: AI has traditionally been perceived as a game of scale—larger models yield better resultsYet, the lessons from DeepSeek reveal that optimal engineering designs should focus on both performance and costA technology only becomes transformative when it is economically viable and easily accessible.

This raises a fundamental question: is the ultimate goal of the AI revolution to serve capital valuation or to enhance human productivity? DeepSeek's value does not rest in triumphing over OpenAI but in affirming that the true measure of technological advancement is not the size of funding or the quantity of GPUs, but rather the reduction in costs and the level of accessibility.

As the haze surrounding capital-constructed AI dissipates, it will be those enterprises capable of translating algorithmic precision into social efficiency that emerge as the winners of the new era