Is DeepSeek a Sputnik Moment?

I'm not sure that "software will eat the world," but it could consume the stock market bubble in a single gulp.

Is DeepSeek a Sputnik Moment? Let's break it down. The Soviet Union's October 1957 launch of the world's first artificial satellite, Sputnik 1, stunned the U.S., which reckoned it had a commanding lead in "the Space Race." (It turns out the U.S. had the capability of launching a satellite before Sputnik, but held off for various reasons.)

That a geopolitical rival had reverse-engineered advances and leapfrogged the U.S. shocked America into a multi-decade response that culminated, at least in the public perception, in America winning "the race to the Moon" by landing the first humans on the Moon in July, 1969 in the Apollo 11 mission.

The shockwaves generated by a Chinese company's release of a suite of AI tools called DeepSeek last week may well rival the Sputnik shock, as the DeepSeek AI tools appear to meet the same benchmarks as AI tools such as those issued by OpenAI and other companies, but requiring far less computing resources.

DeepSeek achieves its capabilities not from expensive hardware (processors) but from advances in software that can be used on smartphones. The software innovations embedded in DeepSeek have profound financial implications for the companies that manufacture the costly processors needed by conventional AI data centers--Nvidia is the dominant chipmaker in this market--and the Big Tech companies spending billions of dollars (called capex in the financial realm, short for capital expenditures) to create AI tools that they can eventually sell via the subscription model.

DeepSeek software evaporates 1) the need for super-energy-hungry, super-expensive processors, 2) vast quantities of electricity and 3) the market for paid subscription AI tools, as DeepSeek's software runs on standard processors and it's been released as open-source software which can be downloaded and run offline on local resources such as PCs or smartphones.

In effect, the AI hardware monopoly and quasi-monopoly of AI software has been broken, and like Humpty-Dumpty, it can never be put back together again. I was skeptical of DeepSeek, but correspondent Cheryl A. sent me this article, which changed my mind. The author is careful to delineate his chops in the fields of AI and investing, and to dismiss this as uninformed is very likely a mistake.

The Short Case for Nvidia Stock

I read the "Theoretical Risks" section carefully and concluded that what the DeepSeek developers did was take the loss of precision performed at the end of conventional AI via compression and move it into the learning / reward process, where it did the work with less precision but with 45X less CPU/memory/cost.

Necessity is the mother of invention: unable to get NVDA chips in big numbers, the Chinese programmers were forced to innovate in software much like programmers on deep-space missions like Voyager 1, which carried extremely limited CPU and memory onboard. DeepSeek's developers made clever use of software to avoid needing super-duper processing power.

Voyager 1, launched in 1977 with three tiny computers packing a mighty 69 kilobits of memory (one low-resolution JPEG photo) in total and 8k per second processing power, is still functioning 47 years later, as programmers worked around a component failure with clever software.

Some of the clever software techniques used by DeepSeek reminded me of the workarounds deployed by the Voyager team last year when the spacecraft stopped responding. It's worth reading this brief account of their workaround:

NASA's Voyager 1 Resumes Sending Engineering Updates to Earth:

The team discovered that a single chip responsible for storing a portion of the FDS memory -- including some of the FDS computer's software code -- isn't working. The loss of that code rendered the science and engineering data unusable. Unable to repair the chip, the team decided to place the affected code elsewhere in the FDS memory. But no single location is large enough to hold the section of code in its entirety.

So they devised a plan to divide the affected code into sections and store those sections in different places in the FDS. To make this plan work, they also needed to adjust those code sections to ensure, for example, that they all still function as a whole. Any references to the location of that code in other parts of the FDS memory needed to be updated as well.

The team started by singling out the code responsible for packaging the spacecraft's engineering data. They sent it to its new location in the FDS memory on April 18. A radio signal takes about 22 1/2 hours to reach Voyager 1, which is over 15 billion miles (24 billion kilometers) from Earth, and another 22 1/2 hours for a signal to come back to Earth. When the mission flight team heard back from the spacecraft on April 20, they saw that the modification worked: For the first time in five months, they have been able to check the health and status of the spacecraft.

I was also skeptical of the possibility--suggested to me by Adam Taggart, host of Thoughtful Money--that DeepSeek could conceivably prove to be the Black Swan that pops the AI bubble. This now strikes me as less of a possibility and more of an inevitability. There is simply no longer enough advantage generated by super-energy-consuming, costly chips in terms of generating a product that is worth paying for when equivalent tools are already available for free that can run offline on free-standing devices--which means there can't be any back-door stealthy "calling home" by the software.

And given that the Mag 7 stocks, most of which are heavily invested in AI, are almost 40% of the total market capitalization of the U.S. stock market, and the global concentration of equity wealth in the U.S. market, then the AI bubble popping will pop the entire global stock market bubble--and with that reduced to rubble, the entire global Everything Bubble in risk-on assets.



The potent phrase software is eating the world may manifest in ways AI investors did not reckon possible when they projected billions of dollars in high-margin profits from AI chips and tools. Whatever we think of DeepSeek's approach, others will quickly reach onto its open-source toolbox and develop their own innovations based on its core structures.

We can be sure that the smart money has already overheard the discussion between Dave and HAL 9000 about the need for power-hungry chips and stiff monthly fees for AI tools:



I'm not sure that software will eat the world, but it could consume the stock market bubble in a single gulp.




New podcasts:

KunstlerCast417: Charles Hugh Smith, Progress and Anti-Progress (1 hour)

Charles Hugh Smith on the Extremes in the U.S. Economy and Markets. (26 min)

---

My recent books:

Disclosure: As an Amazon Associate I earn from qualifying purchases originated via links to Amazon products on this site.

The Mythology of Progress, Anti-Progress and a Mythology for the 21st Century print $18, (Kindle $8.95, Hardcover $24 (215 pages, 2024) Read the Introduction and first chapter for free (PDF)

Self-Reliance in the 21st Century print $18, (Kindle $8.95, audiobook $13.08 (96 pages, 2022) Read the first chapter for free (PDF)

The Asian Heroine Who Seduced Me (Novel) print $10.95, Kindle $6.95 Read an excerpt for free (PDF)

When You Can't Go On: Burnout, Reckoning and Renewal $18 print, $8.95 Kindle ebook; audiobook Read the first section for free (PDF)

Global Crisis, National Renewal: A (Revolutionary) Grand Strategy for the United States (Kindle $9.95, print $24, audiobook) Read Chapter One for free (PDF).

A Hacker's Teleology: Sharing the Wealth of Our Shrinking Planet (Kindle $8.95, print $20, audiobook $17.46) Read the first section for free (PDF).

Will You Be Richer or Poorer?: Profit, Power, and AI in a Traumatized World
(Kindle $5, print $10, audiobook) Read the first section for free (PDF).

The Adventures of the Consulting Philosopher: The Disappearance of Drake (Novel) $4.95 Kindle, $10.95 print); read the first chapters for free (PDF)

Money and Work Unchained $6.95 Kindle, $15 print) Read the first section for free


Become a $3/month patron of my work via patreon.com.

Subscribe to my Substack for free

---

NOTE: Contributions/subscriptions are acknowledged in the order received. Your name and email remain confidential and will not be given to any other individual, company or agency.

Thank you, David B. ($70), for your marvelously generous subscription to this site -- I am greatly honored by your support and readership.		Thank you, William T. ($70), for your superbly generous subscription to this site -- I am greatly honored by your support and readership.

Thank you, Evan C. ($70), for your magnificently generous subscription to this site -- I am greatly honored by your support and readership.

Thank you, Peter U. ($7/month), for your splendidly generous subscription to this site -- I am greatly honored by your support and readership.