News

Laurent Denize – Global Co-CIO ODDO BHF

On the morning of October 4, 1957, the world awoke to shocking news: the Soviet Union had beaten the United States in the race to space. The surprise was so profound that the launch of the world's first artificial satellite, Sputnik, became a defining moment in history. Now, some are calling DeepSeek’s emergence China’s own "Sputnik moment" for the artificial intelligence ecosystem. To what extent does this comparison hold? Is DeepSeek a turning point for the AI industry? Does it fundamentally reshape the AI investment landscape? Is it altogether a “deep sink” for U.S. exceptionalism and a “deep boost” for China? Rather than relying on ChatGPT and DeepSeek for analysis, let’s use our own reasoning to measure the impact of this Chinese AI breakthrough on asset allocation and investment positioning.

DeepSeek, the Chinese startup shaking the AI world

Founded in 2023, DeepSeek is a Chinese startup that has quickly made waves in the AI landscape. Very recently, it released two open-source ChatGPT-like Large Language Models (LLMs): DeepSeek-V3 in December 2024 and DeepSeek-R1 in January 2025. These models leverage several breakthroughs in architecture, achieving performance on par with current frontier models while significantly reducing training compute costs.

Headlines suggest that the V3 model was trained in just two months using approximately 2,000 Nvidia H800 chips at an estimated cost of $5.6 million - a fraction of the hundreds of millions spent by OpenAI, Google, or Meta on their leading AI models. Is that claim true? It seems most industry experts acknowledge DeepSeek’s performance and 10x efficiency gains. Many think DeepSeek’s methodology is legitimate and that such cost savings are plausible.

In any case, we believe DeepSeek challenges the long-standing assumption in the AI community that "bigger is better", which has been driving the AI trade for the past decade.

Macro Boost? Macro Drag? Micro Boost? Micro Drag?

On the macro side, the primary near-term risk to GDP is that more efficient model training and declining compute costs could reduce AI-related infrastructure capital expenditure (Capex). This impact could be particularly pronounced in the U.S., where some estimates suggest a drag of 0.1–0.2 percentage points on GDP.

However, hyperscalers Microsoft, Meta, Amazon and Alphabet have recently reaffirmed their commitment to continued spending. Therefore, a new, tested AI technology with significantly lower development costs could, over the medium term, reshape the current competitive landscape and create macroeconomic upside, in our view.

In such a scenario, the benefits of AI would spread faster and at a larger scale across the global economy, accelerating adoption and driving productivity gains. From a global macro perspective, this dynamic should translate into higher economic growth.

Additionally, falling AI costs driven by AI commoditisation could act as a deflationary force in the global economy. Against this backdrop, we maintain a long-duration stance in the U.S. until productivity gains fully permeate all sectors. However, with productivity gains likely to have a positive impact on U.S. growth, we may need to revise our expectations for neutral interest rates… unless disinflation turns into deflation. One key factor to monitor is China’s ongoing deflationary pressures linked to its overcapacity, which have yet to fully play out.

On the micro side, we believe incrementally lower AI adoption costs could accelerate productivity cycles, benefiting profit margins not only for U.S. stocks but also for AI-driven companies across Europe and beyond. Additionally, we anticipate a potential consumer product upgrade cycle, where devices such as smartphones and automobiles increasingly leverage AI capabilities. However, this shift may compress profit margins for current tech leaders like Nvidia, Alphabet, Meta or Microsoft.

Not a “DeepSink” for the Magnificent-7, but think diversification

DeepSeek has served as a stark reminder of the hyper-concentration in indexes and the disproportionate positioning in just a handful of U.S. stocks. The DeepSeek breakthrough leads us to three important observations.

First, increased AI competition could drive higher tech spending, but not necessarily to the benefit of today’s AI winners. Reduced Nvidia hegemony could favour better portfolio diversification and enhance alpha generation. When the world's largest stock drops 16% in a single day, concentration risk becomes painfully evident.

Second, AI-related Capex by mega-cap tech companies could become somewhat obsolete if cheaper solutions emerge. In essence, the moat protecting the Magnificent-7 may not be as wide as previously thought.

Third, for the first time, questions are emerging about how long the U.S. can maintain its AI leadership.

Ultimately, our conviction is that this disruption represents more of a rotation story than a bearish view on artificial intelligence. However, we must remain vigilant, as the sky-high valuations of the Magnificent 7 stocks, with a market-cap weighted P/E of 35x, leave little room for disappointment.

Time for Adopters

The arrival of DeepSeek is a catalyst for navigating the AI landscape differently and rethinking AI opportunities across sectors.

• Cautious on Semiconductors and Infrastructure:DeepSeek casts doubt on the demand for expensive high-end chips, as intensifying competition could impact chip manufacturers. Additionally, its model is reportedly 10x more energy-efficient than ChatGPT and other AI technologies, which may weigh on energy demand. However, we highlight four counterpoints:

1. Lower costs could accelerate the rate of AI adoption.
2. DeepSeek may not be easily transferable to all existing ecosystems.
3. Hyperscalers have not slashed their Capex programs.
4. Data centres are only one driver of higher power demand, with decarbonisation remaining a critical priority.

Thus, DeepSeek’s emergence does not jeopardise electrification investment opportunities or our strong conviction in the energy transition.

• Rather Positive on Hyperscalers: We see a mixed impact on hyperscalers. While they are consumers of their own models, companies like Amazon and Google also sell access to both internal and external models. DeepSeek’s greater efficiency and lower capital intensity could reduce operating expenses and Capex in the medium to long term, as Amazon, Google and Meta likely adopt its innovations. Regarding top-line risk, Meta appears best positioned, followed by Amazon and Google. Meta does not generate material revenue from access to Llama; Amazon relies on a range of external model developers (including Anthropic) while Google uses its own model, Gemini, as well as third-party models. In the short term, we do not expect hyperscalers to scale back investments, as demand for AI continues to grow. Most hyperscalers remain capacity-constrained, with AI-driven backlogs far exceeding their Capex ramp. This is reassuring. However, questions around Capex and return on invested capital (ROIC) are unlikely to disappear and must be closely monitored. Lastly, we see a competitive edge for the open-source approach compared to proprietary AI models. Open-source AI fosters rapid success by making cutting-edge technology widely accessible, encouraging global collaboration, reducing costs, enhancing transparency, and intensifying competition with proprietary models that have long dominated the landscape.
• Positive on AI Adopters: As AI becomes more affordable, we believe companies will adopt it more quickly and widely. This is positive for the many potential productivity and revenue beneficiaries of AI technology across most sectors in the stock market, including Industrials, Financials, Internet. Webelieve this acceleration in adoption could reinforce Software as a notable outperformer.

China will benefit from the structural shift in the AI landscape

For years, at least in the collective imagination, China has struggled to match the US in AI development. Early efforts, like Baidu’s chatbot, failed to meet expectations. However, DeepSeek’s success has reshaped the narrative, proving that China can outcompete - if not outperform - established AI U.S. giants. In this respect, the conclusions of the “Draghi report on EU competitiveness”, published in September 2024, have never been more relevant. The report brilliantly underscores China’s lead over the U.S. and Europe in most complex digital technologies, ranging from the Internet of Things (IoT) to AI, cloud computing and quantum computing.

Investors who favour leading companies with strong moats cannot ignore that it is Chinese firms, not Western ones, that now boast the widest and deepest moats, alongside their economic superiority.

We believe that increasingly agile Chinese models and significant improvements in computing cost efficiency, spearheaded by DeepSeek, could drive broader adoption of AI applications. This will pave the way for further global expansion of Chinese players, assuming they can continue to access computing chips.

Beyond attractive valuations (the Hang Seng China Enterprises Index trades at a 12-month forward P/E below 10x), the Chinese opportunity lies in its high exposure to soft technology sectors, such as the internet, software and services. Representing 32% of the market capitalisation and 37% of earnings in the offshore index, these sectors are directly exposed to AI longer-term productivity benefits.

“Sputnik Moment” is a new market exaggeration

For those watching the global AI race, DeepSeek’s development is a reminder that innovation doesn’t solely come from established giants. It can emerge from anywhere. No one can assume a permanent and consistent edge, and changes in leadership are a real possibility. However, just as Sputnik catalysed the Apollo Program and ultimately reaffirmed U.S. technological dominance, we see DeepSeek’s emergence as a wake-up call for the U.S. (and perhaps the European) AI industry. This development is likely to inspire a new wave of innovation in AI. That said, the U.S. is not fundamentally lagging, as it was in 1957, but rather must adapt to shifting dynamics. In this respect, the Sputnik analogy diverges. In the end, could DeepSeek prove to be an unexpected gift from China to the U.S.

Disclaimer

This document has been prepared by ODDO BHF for information purposes only. It does not create any obligations on the part of ODDO BHF. The opinions expressed in this document correspond to the market expectations of ODDO BHF at the time of publication. They may change according to market conditions and ODDO BHF cannot be held contractually responsible for them. Any references to single stocks have been included for illustrative purposes only. Before investing in any asset class, it is strongly recommended that potential investors make detailed enquiries about the risks to which these asset classes are exposed, in particular the risk of capital loss. 

ODDO BHF 

12, boulevard de la Madeleine - 75440 Paris Cedex 09 France - Phone: 33(0)1 44 51 85 00 - Fax: 33(0)1 44 51 85 10 –

ODDO BHF SCA, a limited partnership limited by shares with a capital of €73,193,472- RCS 652 027 384 Paris –

approved as a credit institution by the Autorité de Contrôle Prudentiel et de Résolution (ACPR) and registered with ORIAS as an 

insurance broker under number 08046444. - www.oddo-bhf.com