Does AI really have a $600 billion problem?

One of essentially the most interesting conversations at first of the second half of the 12 months was an article published by Sequoia Capital earlier this week. This was the $600 billion AI challenge.

In short, it advocates selling infrastructure to realize economic performance far beyond what we’ve now. In other words, we’re on an extended and winding road in AI history that goes far beyond the fast cash-out story involving only NVIDIA Corp. and nobody else.

As we follow the rapid evolution of AI from chips to customer experience, nearly every category within the technology industry ecosystem is being touched and radically transformed.

The golden road to riches began with NVIDIA, which has reached a staggering market value in a brief time period. But that is not all. There is a significant digital transformation going down, which I prefer to divide into three major phases. This is what I call the multi-level network effects of AI beyond GPU and chip manufacturing (e.g. Taiwan Semiconductor Manufacturing Company), starting from internal infrastructure consumption to the sell-off of access to computers and software for AI.

The first network effect involves servers, storage, networking and memory from Dell Inc., Micron, Marvell Technology Inc., Broadcom Inc. and others. These firms provide the infrastructure for the enablers. (For devices, these are OEMs, but a distinct set of chips provided to device makers. Qualcomm Inc. and Intel Corp. will join Advanced Micro Devices Inc. and NVIDIA.)

The second network effect shifts to platform providers and independent software vendors, including hyperscalers and SaaS akin to Amazon.com Inc., Alphabet Inc.’s Google, Microsoft Corp., Salesforce Inc., ServiceNow Inc., Oracle Corp., OpenAI, SAP and others. These players are constructing infrastructures for industries.

The third network effect involves industries that use AI to create recent and improved experiences, from retail to financial services. This is where the best consumption of infrastructure will occur and where the best economic profit will accrue as we talk in regards to the multi-trillion dollar impact of AI by the top of the last decade.

Indeed, it isn’t hard to see the concerns about AI sell-outs and associated consumption across industries. Where are AI and generative AI getting used by banks, hotels, restaurants, manufacturing, and other industries – and which firms are creating wealth from these AI-powered advances?

This is what Sequoia meant when it denounced the $600 billion AI dilemma. Then again, perhaps most of those alarmist statements have more to do with the time horizon of AI than with the long-term viability of its expansion.

Amid the recent wave of enterprise capital and equity research reports from Goldman Sachs, Sequoia, a16z, one specific example from Goldman stood out: the power of AI to outperform humans at certain equity evaluation and model constructing. But Jim Covello, head of Goldman Sachs Equity Research, identified that the appliance costs about six times as much human capital.

However, a lot of the AI ​​buildout, representing greater than 50% of capital spending, is coming from a small variety of hyperscale cloud providers and huge enterprises that use AI to support their businesses. And the GPUs will not be exclusively for GenAI, but for model constructing, training and deployment for suggestion engines, video rendering, content filtering and more.

These are lots of the “legacy AI” applications, and they’re going to proceed to proliferate. The same firms which have long benefited from AI proceed to speculate in AI adoption and democratize access to expensive AI infrastructure needed to innovate across industries. Still, these applications are still of their infancy and can likely take several years to turn into widespread and fully measurable.

Meanwhile, this waterfall of AI spend, which starts with power, heat, design, silicon, and manufacturing, flows to systems like compute, storage, and networking, then flows to OEMs/ODMs, after which to ISVs and consultants who support implementation.

The proliferation of AI will take months and years, not days and weeks. But the claim within the Goldman note that there aren’t any killer-gen AI applications is solely false. Use cases for promoting, search, content creation, video development, rendering, and even private models that may streamline financial services or personalize healthcare are evolving quickly, and – unlike the dot-com boom – these apps are here to remain and can fuel market growth.

To be clear, I’m not saying there aren’t legitimate concerns in regards to the time horizon and required return on the tons of of billions of dollars in capital spending.

But this technology shouldn’t be cyclical. It is transformative and can change every industry. Some stocks could also be ahead of the market, others may lag behind by way of real AI-driven value.

Companies and investors who see the long-term impact of AI know that, at worst, that is short-term, front-loaded capital spending that may almost definitely determine the winners and losers of an AI-powered future.