71. The Big GPU Short, China Chips Reach-around, Explaining Broadcom VMware, and CrowdStrike Reset

In this episode of theCUBE Pod, theCUBE analysts John Furrier and Dave Vellante explore significant developments in the tech world, including the launch of the NYSE CUBE East Studio, expanding theCUBE’s reach across major markets.

New episodes every Friday. Subscribe for weekly tech analysis.

Also, during this week’s podcast, Furrier and Vellanet delve into concerns about a potential "big GPU short," drawing comparisons to the housing crisis, and discuss the Broadcom-VMware merger's impact on customers and the broader tech ecosystem.

The episode also touches on U.S.-China tech tensions, particularly China's circumvention of U.S. chip bans, and provides sharp political commentary on Kamala Harris's unrealized capital gains tax proposal and its potential ramifications for the economy.

This Week in Enterprise:

Beyond OpenAI: The rise of not-too-large language models

A flurry of new artificial intelligence models this week illustrated what’s coming next in AI: smaller language models targeted at vertical industries and functions.

Both Nvidia and Microsoft debuted smaller large language models too. Also supporting the notion of more customized models — call them VLMs — OpenAI made its GPT-4o fine-tuning generally available. As much as LLMs have captured much of the attention, these smaller, more controlled models look appealing to enterprises concerned about data governance and privacy, not to mention efficiency.

Indeed, Chinese startups are heading in the same direction, partly to save energy and partly to avoid the need for the most advanced Nvidia graphics processing units to which they don’t have access under export controls. That said, it looks like many Chinese companies are getting access to that high-end computing power through cloud providers such as Amazon Web Services.

Advanced Micro Devices CEO Lisa Su doubled down this week on her quest to slice off a chunk of Nvidia’s lucrative GPU market, as it acquired AI infrastructure provider ZT Systems.

Infrastructure observability firms are having a moment. Not too long after Cisco Systems closed its acquisition of Splunk, others continue to reap the rewards, including Datadog turning in an upside quarter earlier this month. This past week, Grafana Labs raised a boatload at a $6 billion valuation.

Snowflake shares dropped almost 15% Thursday after a disappointing revenue outlook as well as concerns about profitability. But everyone else had pretty positive earnings reports, including Palo Alto Networks, Workday, Synopsys, Zoom and Zuora.

Autonomy founder Mike Lynch sadly died at sea off Sicily with several others, celebrating just a couple months after winning his long-running HP court case. Oddly, co-defendant Stephen Chamberlain was hit by a car and died earlier this week.

TheCUBE and theCUBE Research analysts will be at VMware Explore Monday through Wednesday to suss out what’s happening with the virtualization and cloud pioneer under new owner Broadcom. Also next week: earnings reports from more bellwethers such as Nvidia, Salesforce, CrowdStrike, Dell, NetApp, Pure Storage, HP, MongoDB, HashiCorp and more.

People mentioned in this podcast:
Michael Burry, American investor
Mike Lynch, founder of Invoke Capital
Leo Apotheker, former CEO at HPE
Lina Khan, chair of the FTC
Kamala Harris, 49th United States vice president
Elizabeth Warren, U.S. Senator
Joe Kernan, news anchor
Jeff Bezos, chairman of Amazon
Steve Jobs, co-founder and former CEO and chairman of Apple
Britney Spears, American singer
Eric Schmidt, former CEO and chairman of Google
Ryan Gosling, Canadian actor
Steve Carrell, American actor and comedian
Matt Garman, CEO at AWS
Cole Crawford, founder and CEO of Vapor IO
Jensen Huang, founder and CEO of Nvidia
Lisa Su, chair and CEO of AMD
Pat Gelsinger, CEO of Intel
Steve Chen, co-founder and former CTO of YouTube
Donald Trump, 45th U.S. president
Ronald Reagan, 40th U.S. president
Michael Saylor, co-founder and executive chairman at MicroStrategy
David Floyer, analyst emeritus at theCUBE Research
Rob Strechay, managing director and principal analyst at theCUBE Research
Hock Tan, president and CEO of Broadcom
James Watters, senior director of R&D, Tanzu Division, at Broadcom
Trinity Chavez, lead anchor at NYSE
Brian J. Baumann, director of capital markets at NYSE

#SiliconANGLE #theCUBE #theCUBEPod #theCUBEResearch #NYSECUBEEastStudio #TheBigGPUShort #TechMergers

[michaelwu]

I would have liked to hear a more detailed construct of the GPU crash scenario. Backup, data, etc and not just hearsay.

[daveunbranded]

I wish someone would talk about specific workloads that are not being done now that can be done better.

[furtsmagee]

The cube will sign a $150 million podcast exclusivity deal with Broadcom audio

[EightSox-lq]

Why can we find better models in D Wave tech. Would any new technologies be applied other than what has been described on this Cube?

[dimitargueorguiev]

I remember the HP Autonomy fraud scandal - it was big deal, impacted badly HP and I guess precipitated the HP split. BTW, this GPU short is caused chiefly by the first Gen (that is, inferior) AI models (mainly LLM/Transformers) which require immense resources for training. Think of the 60s of the previous century when the US auto market was flooded with gas guzzling cars which worsened the impact of the oil shortages. I think the situation with GPU demand will be quite different when the next gen AI models for NLP tasks appear. On the other topic: we do not need TikTok, Temu, Shein and the likes; unfortunately, these entities are being cajoled in the US market by the Big Tech which rely on ads - Alphabet and Meta. On the VMWare topic - I hope Broadcom is not Thoma Bravo..

[emmettobrien]

GPU crash?? Huh? Most LLM ‘s are taking 12-18 months to train (on GPU’s), when released they are already outdated by 18 months. Then everyone has to use GPU’s to “refine” the LLM’s (mostly redundantly on the same public data) to catch them up to date with specific data interests. Then there is the massive backlog of about 2 petabytes per enterprise that need to use GPU’s to convert their data into vector databases so AI LLM’s can be used by enterprises to access their own data, privately/securely. The limiting factor for the US, assuming the GPU demand can continue to be met, the US does not have enough electrical power generation capacity to meet the never ending amount of AI LLM training, re-training, and enterprise vectoring and refining backlog. We need to start building cross border power distribution from Canada now. We can manufacture and meet the needed GPU demand but do we have the political will to meet the required corresponding power generation demand?

[fcaspergerrainman]

The key thing that literally no one talks about, is education gap/ skills gap, and most people in their 30s and 40s around the world are addicted to their social media, so most office workers just don’t function properly. The AI level, esp Gen AI can increase productivity in the work place (see JP Morgan 2024 investor day transcript), but most CEOs don’t understand tech, and most workers are too stupid and lazy to reskill and upskill. It’s a human problem not AI problem.