Between the Devil and the Boundless AI Datacenter

It’s a Thursday. No more needs to be said. Let’s have a look at some of the developments that have made the grade as intriguing news items.

As a public backlash to AI grows, the captains of the AI industry belatedly recognize that they ought to dial down their prophecies of labor-force armageddon. What do they really believe, though? Is what they said previously an accurate reflection of their assessments of AI’s malign influence on human employment, or is their revised position an honest, more modest reappraisal of AI’s impact?

We cannot know the thoughts flitting through the minds of the industry’s CEO class, but we have reason to question the honesty of their public assertions. The grim truth is that labor-force diminution had become a selling point for AI. Unfortunately, those at the pointy end of AI’s labor displacements are apt to respond unfavorably to the technology’s workplace incursions. AI CEO’s might have to resort to dog whistles, coded messages, and secret handshakes when pitching their technology’s to their cost-cutting counterparts in other industries.

Quarterly Results: Where Success and Failure Walk Hand in Hand

In technology’s public markets, the principal question is no longer “what have you done for me lately?” A company can report impressive quarterly results, as Zscaler did this week, but if investors perceive sales guidance to be lacking in vigor, look out below. In the case of Zscaler, the guidance wasn’t at all egregious, as this extract from Barron’s attests:

Looking ahead to the fiscal fourth-quarter, Zscaler expects revenue between $875 million and $878 million, coming in below Wall Street’s $878.6 million estimate. The company sees fourth-quarter profit between $1.08 and $1.09 a share. Analysts peg fourth-quarter earnings at $1.03 a share, according to FactSet.
For the full-fiscal year, Zscaler forecasts earnings of $4.10 and $4.11 a share with revenue between $3.32 billion and $3.33 billion. That is up from its previous expectation of $3.99 to $4.02 a share with sales totaling $3.31 billion to $3.32 billion.

If that sort of guidance is considered disappointing, the margin for error is wafer thin. The forward-looking revenue number, even if it comes in at the low end of Zscaler’s expectations, falls only $3.6 million shy of the analysts’ consensus. Meanwhile, Zscaler’s projected profit surpasses analysts’ expectations. The company’s forecast for full-year fiscal results is similarly solid. Nonetheless, Zscaler shares were hammered, dropping more than 21.5%, in pre-market trading Wednesday.

At least some of the apparent market overreaction is attributable to — you guessed it — AI, which is widely perceived as an angel of death hovering menacingly above the headquarters of software companies in Silicon Valley. Still, until more evidence of AI hegemony is available, we must continue to apply the word “overreaction” to the market's hypercritical reactions to otherwise commendable financial results.

AI Costume Party: Vices Dressed as Virtues

In business as in life, we occasionally feel compelled to portray our vices as virtues, to frame our mistakes, even our worst strategic missteps, as visionary masterstrokes. Such is the case at Meta, where CEO Mark Zuckerberg now positions his company’s extravagant outlays for AI datacenters as a potential prologue to becoming an IaaS cloud provider. Zuckerberg discusses this possibility as if it were part of the initial plan. The truth, of course, is that Zuckerberg is scrambling to transform hulking, energy-sucking lemons into intoxicating spiked lemonade.

Unlike the Iaas cloud giants — AWS, Google, and Microsoft Azure, and Baidu, Alibaba, and Tencent (BAT) in China — Meta (formerly known as Facebook, before Zuckerberg plunged that company into the fetid metaverse) has neither history nor credibility as a B2B player. If Zuckerberg were to sell excess processing capacity to enterprises, the move would amount to a desperate effort to compensate for AI overspending. Any move into the IaaS space by Meta will be a tacit concession of a significant tactical error, a recognition that the company braintrust grossly overestimated AI demand and revenue generation from its core consumer market. The consumer business and IaaS are very different propositions with different requirements and customer bases. Meta’s move into IaaS would require a separate company, similar to what Amazon did when it created AWS, and the cost would be prohibitive — establishing datacenters in every major center of global commerce — if Meta were serious about becoming a credible player.

Success would be far from assured, failure more likely. That’s why I think this is nothing more than a feint by Zuckerberg, an act of misdirection designed to assuage nervous investors while Meta anxiously waits for its core business to achieve more AI traction.

Rinse and Repeat

If you spend enough time in this industry, you’ll see that cycles and patterns really do repeat themselves, sometimes with only the slightest variation. Less than a decade ago, Dell and HPE led the charge for private and hybrid cloud, running cloud workloads in whole or in part in on-premises infrastructure. From there, Dell and HPE launched into variations on the theme, including cloud-related edge computing, repatriation of workloads from the cloud to on-premises enterprise datacenters and edge environments, and the imperative of sovereign cloud.

Past is prologue. That’s from William Shakespeare’s The Tempest. Methinks the Bard would be unsurprised to discover that Dell is now talking about private and hybrid AI. It’s basically the same pitch Dell invoked in proselytizing private and hybrid cloud, except now it’s aimed at proliferating AI tokens. It’s a well-worn playbook, replete with plays Dell can run from decade-old muscle memory. The challenge, as before, is not only in the practiced positioning but in the execution.

Dell must ensure that the on-premises option is as agile, flexible, scalable, and efficacious as anything that enterprise customers can obtain from deep-pocketed and endlessly resourceful cloud providers. I don’t blame Dell for reprising this strategic gambit— selling on-premises AI infrastructure is a big part of what the company does — but the figurative devil is always in the details.

Speak of the AI Devil

Speaking of the devil, let’s consider one of his nemeses: the pope. Why are we making the Vatican a port of call? Well, we’re making the jaunt because the pontiff issued an encyclical letter earlier this week that amounted to a Festivus-like airing of grievances against AI. The grievances were extensive, delivering yet another public-relations setback to a technology that doesn’t need the bad press. After taking aim at the environmental impact of AI, particularly its voracious energy and water appetite, the pope also warned of “new forms of slavery” engendered by the need for minerals in the tech supply chain. He also cited the threat of dehumanization, deriving from the growing supremacy of technology.

Quoting the pope, as reported by Barron’s:

“If technology becomes the ultimate criterion, the human person risks being reduced to data, a cog in a machine or a commodity,” he wrote.

You don’t have to be a Catholic, or an adherent of any major organized religion, to grant validity to that concern. It’s a sentiment that the industry should take seriously and address honestly, which brings us back to the first item of today’s post.

AI has an image problem, partly the creation of the industry’s major players. Candor is required, but we seem to have entered a period where “authenticity” has devolved into performative affectation.