Big Tech Hiring Is Rotating Toward AI Infrastructure

Tae Kim characterizes the technology hiring market as a thaw rather than a return to the broad expansion of the low-rate era. Big companies are hiring again, he says, especially in technology, but the relevant demand is selective: AI-specific roles, core engineering, silicon, and infrastructure.

That distinction is central to his account. Kim contrasts the current market with what he calls the ZIRP-era hiring craze, when companies could hire “thousands of product managers” partly to keep them from competitors. The reopening he describes is not broad-based defensive staffing. It is a competition for capabilities that companies need if they intend to build and train large models.

The two hiring visuals shown during the discussion describe the recovery at different levels. A Bureau of Labor Statistics chart tracks tech-sector job openings from the first quarter of 2022 through the first quarter of 2024. Its line trends upward in the final quarter shown. A separate chart attributed to TrueUp.io follows job openings at Meta, Amazon, Alphabet, and Microsoft from 2020 through 2024: openings fall steeply in 2022, then recover gradually in late 2023 and early 2024.

Neither graphic breaks openings down by job category. Kim supplies that interpretation himself: the renewed pressure is strongest around the technical functions that make AI development possible. He names silicon and AI infrastructure alongside core engineering, and says companies are prepared to pay for people with those skills.

A post shown on screen under Kim’s @firstadopter handle makes the same claim in compressed form: “The narrative that tech is dead is wildly inaccurate. Big tech is quietly accelerating hiring for silicon and AI infra roles. It’s a rotation, not a recession.”

“Rotation” distinguishes his argument from a general claim about technology employment. The work being valued has shifted toward the systems beneath the model: the hardware-facing expertise, infrastructure, and engineering capacity required to train and operate large-scale AI efforts. Kim points to Nvidia as drawing substantial talent, but his larger point concerns the traditional software companies as well. They cannot simply leave their own infrastructure organizations paused while pursuing massive-model training.

In Kim’s framing, this is an operating constraint rather than an abstract vote of confidence in the labor market. Companies may have cut staffing heavily in the preceding period, but model-training ambitions create work that cannot be deferred indefinitely. The relevant teams must be built, maintained, or rebuilt; otherwise the organization lacks the capacity to support the strategy it says it wants to pursue.

That is why he places core engineering alongside explicitly AI-focused jobs. The demand is not limited to a narrow set of people labeled “AI.” Training large models depends on a broader technical foundation. Silicon talent matters because computation matters; infrastructure talent matters because models have to be trained and run; core engineering matters because those systems have to work as part of a company’s operations.

The displayed charts give the claim a visible employment-market backdrop: openings first contracted and later recovered. Kim’s emphasis is on the composition of that recovery. The companies reopening requisitions are not, in his description, recreating the earlier practice of adding undifferentiated headcount. They are rebuilding the capacity most directly attached to AI development.

HBO’s short-form experiment addresses a different operational problem: attention to its shows is already occurring in feed-based environments outside its own service. Kim says Warner Bros. Discovery is looking at engagement and churn while watching TikTok and YouTube Shorts absorb viewing around clips from Succession and The Sopranos. People consume “millions of hours” of those clips on free platforms, he says, and HBO does not directly monetize that engagement.

The proposed response is a vertical-video feed inside the company’s app. Rather than presenting only the familiar horizontal rows of films and series, the app would have a dedicated tab for rapid-fire clips. A viewer who finds a scene compelling could use a button in the feed to begin the full episode.

The product shown on screen is not merely a conventional trailer page or a social-media promotion. An HBO-attributed mobile screenshot depicts a vertical feed playing a scene from The Sopranos. It includes like and share controls, with the visible text “HBO Shorts,” “Swipe up for more,” and “The Sopranos Season 1.” The visual borrows the basic interaction model of short-form social video: an upright frame, fast movement between clips, and lightweight engagement controls.

The catalog remains long-form. Kim describes HBO as slicing prestige dramas into bite-sized clips, not replacing those dramas with short programming. The intended function of the clip is discovery: it gives a prospective viewer a moment of interest and then offers an immediate path into the full episode.

A TBPN post shown on screen presents the strategy in one sentence: “HBO is officially testing short form vertical video to capture Gen Z watch time.” A separate browser screenshot attributed to Variety carries the headline, “HBO Max to Introduce Short-Form Video Feed to Compete with TikTok.”

Kim’s explanation puts more weight on the conversion path than on the resemblance to TikTok. On an external platform, a viewer can encounter and share a dramatic scene without ever arriving in HBO’s product. The scene may build cultural presence for the show, but the next action is not controlled by the service. The in-app feed changes that sequence. The viewer can move from a clip to the episode without leaving the interface, searching separately for the title, or deciding later whether to start it.

That phrase joins two aims that Kim identifies separately through engagement and churn. For acquisition, a compelling moment from an established show can become an entry point for someone who has not begun the series. For retention, the same feed gives existing users another way to encounter the service’s catalog and begin watching. In both cases, the product design seeks to make discovery more immediate and to keep it inside the company’s own application.

The choice of The Sopranos in the displayed feed is useful because it makes the strategy concrete. HBO is not described as building the feed around new, standalone vertical-video franchises. It is using scenes from programming it already owns or distributes as raw material for a new discovery surface. A clip is valuable not only because it can be watched quickly, but because it can prompt a longer session with the underlying series.

Kim calls the move wild for a premium cable network because the format historically belongs to a different kind of media environment: platforms organized around rapid, continuous, vertically consumed clips. But the source’s description of the feature is more specific than a simple pivot toward short-form video. HBO is adapting that environment’s discovery mechanics to serve a premium long-form catalog.

Kim’s two examples concern different industries, but each is organized around a problem of control.

For large technology companies, the strategic activity is the capacity to train massive models. A company can announce an AI strategy, but it still requires the people who understand silicon, infrastructure, and the engineering systems around the models. Kim’s hiring argument is that these organizations are once again competing for that capacity because they cannot keep it inactive indefinitely and still pursue the work.

For HBO, the strategic activity is not simply viewing. It is the chain from attention to fuller engagement with the catalog. Short clips of HBO shows already circulate through platforms built around feeds, and Kim says viewers spend substantial time with them there. HBO’s test tries to place the clip, the recommendation surface, and the button to start an episode in one environment the company controls.

The symmetry is not that the two businesses are pursuing the same tactic. It is that both are responding to a dependency that has become too important to leave unattended. In technology, the dependency is specialized internal technical capacity. In HBO’s case, it is the path through which viewers discover its programming and decide whether to watch more.

Kim’s language about hiring—“a rotation, not a recession”—captures the first shift. The labor demand he describes has moved toward work directly attached to AI infrastructure rather than returning uniformly across technology functions. His language about HBO captures the second: the feed is a funnel, not a replacement for prestige television. The vertical clip is meant to create an episode start.