ABSI - The AI Revolution Shaken

Every Tuesday afternoon we publish a collection of topics and give our expert opinion about the Equity Markets.

As-Barclay-Sees-It-Barclay-Pearce-Capital-News-Email-1

In June 2024, Goldman Sachs went contrarian on the AI trade with a warning that AI infrastructure was being overbuilt and was too expensive for the benefits produced. These prescient comments may be playing out when unknown Chinese AI firm DeepSeek unveiled its R1 model. The Company claims it is on par with U.S.-developed AI leaders and was built with just 3% of the resources. ABSI this week explores the challenges to the perceived dominance of U.S. companies in the AI race and questions the massive investments poured into AI infrastructure by tech giants.

DeepSeek has thrust itself into the spotlight with the release of its R1 AI model. Cost estimates are not conclusive but there are claims that R1 was built on less than US$6 million in training hours and 50,000 NVIDIA H800 GPUs. The model runs queries at US$0.14 per million tokens compared to OpenAI’s US$7.50/million tokens.

The implications of DeepSeek's claims are nothing short of revolutionary…if they’re true. 

If true, they suggest that the path to advanced AI capabilities may not necessarily require the massive financial outlays that have become synonymous with the industry. This revelation has the potential to democratise AI development, allowing smaller players to compete on a more level playing field. It also means that the first movers who have spent billions on AI infrastructure may have a white elephant on their hands.

 

estimated training cost

Source: Stanford University

 

The market's response to DeepSeek's announcement was swift and severe. NVIDIA, the chip giant that has been riding the AI wave to stratospheric valuations, saw its stock plummet by nearly 17%, wiping out approximately $600 billion in market value. This dramatic sell-off reflects investors' fears that the AI gold rush, which had propelled NVIDIA to a trillion-dollar valuation, might be built on shifting sands. For further context, the amount of value wiped from NVIDIA overnight was the approximate equivalent of Disney, McDonald’s, and L’Oreal market capitalisation combined.

The ripple effects were felt across the tech sector, with other AI-related stocks also taking a hit. Broadcom fell by 17.37%, while tech giants like Microsoft, Tesla, and Alphabet (Google's parent company) experienced significant declines. The tech-heavy Nasdaq Composite index dropped 3.2%, underscoring the broader impact on the technology sector.

 

NVIDIA stock price

Source: Google Finance

 

While DeepSeek's claims are certainly attention-grabbing, they have been met with a healthy dose of scepticism from industry experts and analysts. Many credible analysts are rightfully questioning the cost transparency, the purported hardware requirements, and the technology claims. It is also important to note that DeepSeek’s parent company is Chinese hedge fund High-Flyer. While unconfirmed, it wouldn’t be a stretch of the imagination to believe that High-Flyer shorted NVIDIA stock and is likely profiting from the massive wipeout of value.   

The timing of DeepSeek's announcement, just a day before former President Trump's unveiling of the ambitious Stargate project—a $500 billion plan for AI infrastructure in the U.S.—adds another layer of intrigue to this story. This sequence of events has fueled speculation about whether DeepSeek's reveal was strategically timed to influence U.S. policy decisions or market sentiment.

Moreover, the development raises questions about the effectiveness of U.S. export controls on advanced AI chips. If DeepSeek has indeed achieved comparable results with less advanced hardware, it could undermine the strategy of maintaining U.S. technological superiority through export restrictions.

As the dust settles, the tech industry and investors are left grappling with fundamental questions about the future of AI development:

  1. Efficiency vs. Scale: Does DeepSeek's approach represent a paradigm shift in AI efficiency, or is it a case of comparing apples to oranges?
  2. Investment Strategies: Will this development lead to a reevaluation of massive AI infrastructure investments, or is it merely a bump in the road for established players?
  3. Global Competition: How will this affect the global AI race, particularly the dynamics between the U.S. and China?

While it's premature to declare a revolution in AI development based on DeepSeek's claims alone, the episode serves as a stark reminder of the volatile and rapidly evolving nature of the AI industry. The incident serves as a cautionary tale about the risks of overreliance on a single narrative in the AI space and as the AI landscape continues to evolve, adaptability and critical thinking will be key to navigating the challenges and opportunities that lie ahead.


 

Introducing BPC Wealth Management

 

BPC Wealth Management is dedicated to shaping resilient investment portfolios, empowering you to achieve and sustain your financial aspirations. While the foundation of your portfolio focuses on long-term investments, through BPC, clients will be offered opportunities in equities trading and equity capital markets. This aspect is highly customised, allowing asset flexibility. Discover how our proactive and client-focused approach can help you achieve your financial aspirations by booking your discovery call with James Whelan. 

3-Aug-27-2024-02-02-51-8847-AM

 


We offer value-rich content to our BPC community of subscribers. If you're interested in the stock market, you will enjoy our exclusive mailing lists focused on all aspects of the market.

To receive our exclusive E-Newsletter, subscribe to 'As Barclay Sees It' now.