Blain’s Morning Porridge 27^th Jan 2025: The Judder Moments in Rates and AI – Deepseek!

“The moment you declare something is priceless, someone will undercut you….”

A sure sign of just how nervous markets are is volatility and judder moments – the realisation that things aren’t what you thought they were. Who knew a bond bear market was upon us, and interest rates are unlikely to fall? Who knew the Chinese AI Deepseek can do what OpenAI can do at a fraction of the cost?

Did you feel the judder moments over the weekend? We have 2 potential disruptive tremblors active in markets:

• The first is President Trump and interest rates – the threat in plain sight. While Trump is playing bully-boy trade-economics with nations from Columbia to Denmark, the recessionary effects on the front-lines of industry and commerce are perhaps more significant and longer lasting.
• The second is the No-See-Um threat that’ emerged from the shadows is Chinese AI wunderkid Deepseek. It could reprice expectations across AI.

With the US economy firing on all cylinders the prospects for the scale of interest rate cuts that Trump demands are limited. Opinion on the analyst desks is the Fed will likely hold rates steady – we will maybe learn more from the Fed meeting this week – and even hike if inflationary threats emerge. This is setting Trump and his tax-less/spend-more fiscal policies on collision course with Fed Chairman Jay Powell, who won’t be stepping down from his post (voluntarily) for another 18 months. Trump is ramping up America First rhetoric with little thought as to the recessionary effects the uncertainty his scatter gun tariff-threats and policy statements trigger elsewhere in business decisions. It’s difficult for businesses to plan for expansion when Trump is musing on a 25% tariff hike because their prime-minister has upset him over some triviality.

Meanwhile, Chinese AI firm Deepseek is setting the markets a-judder this morning as folk wonder if it hails a dot.com moment in AI valuations. Tech investor Marc Andreessen is on the wires calling its “AI’s Sputnik Moment”, when America panicked on the thought the Soviets were winning the space race. Back then they went on to spend billions going to the moon. This time it’s different – nations around the globe are looking at how AI can change their economic prospects, and if the Chinese can provide it cheaper, with open code, without the Trump threats… then why not?

In the course of just a couple of months, Deepseek’s LLM (Large Language Model) has astonished AI watchers as broadly comparable in performance to the latest OpenAI products – but at a fraction of the cost. The Chinese firm has been able to achieve the processing of the stupendous amounts of Data in an LLM – despite the restrictions placed on supplying the latest, dedicated AI Chips to China. They’ve done it with cheaper chips and “demonstrating how software ingenuity can offset hardware constraints”, according to the FT.

To catch up on the Deepseek story, here are a couple of links:

WSJ        Silicon Valley is Raving About a Made-in-China AI Model

FT            Asia Tech stocks fall as Deepseek sows doubts about AI Spending

BBerg   Deepseek shakes up Stocks as Traders Fear for US Tech Leadership

There is plenty of scepticism around Deepseek – how can the Chinese have caught up so fast? Have they an undisclosed horde of Nvidia superchips.. or have they really learnt how to do it better and cheaper? The Wall Street Journal reports that Deepseek said its latest LLM cost $5.6mm to train – a fraction of the $100-$1 bln US AI firms expect it to cost! That is raising a host of issues about effectiveness of the vast sums US firms are earmarking for AI development, and what it means for their valuations, if cheaper Chinese products are going to eat their lunch?

Now I do have some ethical and moral issues with AI Chatbots – mainly because it would be easy for me to tell an AI to write it in the Style of Bill Blain – ie, liberally sprinkle with spelling mistakes and bad grammar. I promise myself and Morning Porridge readers never to let an AI write the Porridge, and I will advise readers whenever I cite them.

This morning I signed up for Deepseek and asked it a question: “Will Deepseek “Smash” US AI companies because it is so much cheaper?”

I’ve edited the responses lightly. They seemed designed to allay any fears, yet made one repeated point: if DeepSeek is reliable, as good as, and cheaper than other AI’s – why would you not use it? My conclusion is US AI tech is now in engaged in global competition, and it raises a host of issues for firms including Nividia et al.

DeepSeek gave me AI-level thoughtful comments on AI competition:

• • Performance: The effectiveness, accuracy, and reliability of AI models are critical. If Deepseek offers a cheaper solution but lacks the performance or capabilities of more expensive U.S. AI technologies, it may not gain significant market share.
  • Innovation: U.S. AI companies often lead in innovation, developing cutting-edge technologies that set industry standards. If Deepseek cannot match this level of innovation, it may struggle to compete.
  • Trust: U.S. AI companies benefit from established trust and brand reputation, especially in sectors like healthcare, finance, and defense. Companies and consumers may prefer well-known providers even if they are more expensive.
  • Regulatory Compliance: U.S. companies are typically well-versed in navigating complex regulatory environments, which can be a significant advantage in highly regulated industries.

• • Compatibility: AI technologies need to integrate seamlessly with existing systems and workflows. U.S. companies often have an edge in creating solutions that work well within established ecosystems.
  • Support and Services: Comprehensive customer support, training, and consulting services can be crucial for the adoption of AI technologies. U.S. companies often excel in providing these value-added services.

• • Diverse Needs: Different markets have different needs and preferences. While cost is a significant factor, it is not the only consideration. Factors like localization, cultural relevance, and specific industry requirements can influence market success.
  • Geopolitical Factors: Geopolitical tensions and trade policies can impact the adoption of foreign technologies. U.S. companies may benefit from domestic policies and international alliances that favor their products.

• • R&D Investment: U.S. AI companies often invest heavily in research and development, leading to continuous improvements and new breakthroughs. This sustained investment can help maintain a competitive edge.
  • Talent Pool: The U.S. has a strong talent pool in AI and related fields, which supports ongoing innovation and development. [So Does China.}

• • Strategic Importance: AI is considered a strategic technology, and many countries may prefer to develop and support their own AI industries for national security and economic reasons.
  • Economic Scale: The U.S. has a large and diverse economy that can support a robust AI industry, providing a significant market for domestic AI companies.

Conclusion: While Deepseek’s lower cost could make it an attractive option in certain markets or applications, it is unlikely to single-handedly “smash” U.S. AI companies. The AI industry is highly competitive and influenced by a wide range of factors beyond just cost. U.S. AI companies have several advantages, including innovation, brand reputation, regulatory compliance, and a strong ecosystem, that can help them maintain their position in the global market. However, competition from lower-cost providers like Deepseek could drive U.S. companies to innovate further and improve their cost structures, ultimately benefiting the industry as a whole.

The bottom line is China has repeatedly shown it can compete against the West. Don’t underestimate it undermining the basis of the current “bubble” in AI valuations.

Finally: I will be off skiing from later this week in Whistler – whole of first week of Feb. If any of my Canadian buddies in BC fancy a chatch-up, use billblain@morningporridge.com to get in touch. We are thinking of big lunch with Porridge Fans up in the hills!

Out of time and off to find a day job…

Bill Blain

Author of the Morning Porridge

www.windshift.capital

billblain@morningporridge.com