Generative AI: focus on the consumer and the consumption

last updated:

Generative AI: focus on the consumer and the consumption

by:
last updated:

What precisely the burn rate will likely be for operating large-scale generative AI is unclear to those of us outside the respective organizations. Still, those “cash on hand” figures will be interesting to watch if you're interested in keeping a rough track. They’re a close-to-real-time tracker on what the long-game bet to own the AI market might be costing.

One of the things that being a general nerd about technology, but one with a specific technology focus that pays the rent, is trying to filter out the otherwise interesting noise that enters my purview every week. I’ve learned to my cost in the past as an analyst that if you try and manage too-wide a coverage area, you’ll end up doing nothing but managing the inflow of information, not just in your daily working time, but evenings, weekends,…. sitting up in bed reading posts on Seeking Alpha by the light of your smartphone screen. It’s a lifestyle alright, just not a healthy one.

That said, even though it is on the face of it consumer technology news, my interest was piqued this week by the announcement that Microsoft had hired Mustafa Suleyman and Karén Simonyan (and an unnamed number of colleagues) from AI start-up Inflection to head up “… a new organization called Microsoft AI, focused on advancing Copilot and our other consumer AI products and research”.

It’s interesting for many reasons—not least that Microsoft isn’t acquiring Inflection, taking a stake in the company, or buying any of its IP. It’s not a traditional “acqui-hire” deal where a company is bought not for what it makes but to gain the collective brain trust within its employee base. I have no inside track; we have a good relationship with Microsoft here at Deep Analysis, but not the relationship that gives us that insight.

For me, however, the explicit focus on “consumer products and research” is most interesting. Why? Well, let’s look at why companies that derive substantial enterprise revenues would want to bother with consumers at all.

We don’t share market share

Let’s take at face value that generative AI and its successors will significantly impact the world of work in the next few years. That is probably a defensible statement for all the boosters and detractors who would want to pull the argument to and fro. And it’s mainly defensible because pretty much every software company that supplies enterprises large and small says it will and has new products and a committed roadmap to back that up. Sure, we can pick that apart over a coffee sometime, as there is undoubtedly nuance I’m deliberately skipping, but let’s move on.

The big bet isn’t on whether generative AI et al. will have a significant impact; it’s whether it’s generationally impactful. Again, the consensus is that it is. These generational shifts are not just regular opportunities to grow but rather the chance to grab a significant market share of something that is greenfield. Looking back on a few recent generational computing events, mobile, social, and cloud, each allowed new or refreshed operators to vault into a revenue-generating spot their previous momentum would not have allowed.

For incumbents with significant market shares, this provides both an opportunity and a threat. Your advantageous position of influence as a significant supplier means you have the first strike at the new thing for those who are already paying you, and that helps you head off the threat of someone new elbowing their way into a chunk of the market. As I mentioned recently over at Reworked, that’s already been playing out around AI regulation.

Proven at home and school, used at work.

If we look at two of those incumbents, Microsoft themselves and Google (Alphabet, as is), they’ve both been afforded opportunities through those 21st-century computing events, to varying degrees of success. Again, we can pick over those respective successes and failures another time, but let’s pick one that I think is especially relevant here: productivity software.

Just over 18 years ago, Google launched its first desktop applications suite, which by 2024 is called – after many name and branding iterations – Google Workspace. A cloud-first set of cheap, basic applications up against the desktop behemoth that was Microsoft Office, just surviving to the present day is an achievement, in no small part given Google’s propensity for canning promising, early-life projects for failing to reach a launch trajectory that suggested eight-figure revenue potential.

One way that Google could break the hold that Office applications had over many users was to offer them for free to everyone. For institutions like those in education, where many young people got their first practical hands-on use of desktop computing, Google pushed especially hard to be adopted. Over time, it made a virtue of forging product innovation amongst its consumer audience; indeed, new functionality was often available to that set of users long before it made it into the paid-for, enterprise tiers of the products. 

It’s one of those situations where – from a vendor perspective – in the end, everybody won. Google? Well, my last user statistics were those cited at Google Cloud Next 2023, which stated Workspace had 3 billion users, with 10 million of those paying for the privilege. For Microsoft? As of its Q2FY24 numbers, it reported 400m paying Office365 seats. It’s still far from an unequal fight, especially regarding revenue. But an example of where forging use outside the enterprise can help foster adoption within it.

Revenue, cash, and capex warfare

It’s worth reminding ourselves just how much cash (or “cash on hand” as it’s often called) Google (Alphabet) and Microsoft are currently sitting on. 

For Alphabet as of the end of 2023, that was $110.92 billion—down from a peak of $142 billion in September 2021—and with its entire Google Cloud (including the Workspace business but a whole heap of other revenue streams, too) operating currently generating a shade over $30 billion a year, it’s a reminder how big the rest of the company’s business is by comparison (total revenues are in excess of $300bn as of 2023). As of the end of 2023, Microsoft held $80.98 billion in cash, down from a peak of $143.95 in September 2023.

The tl;dr here is that both generate a tidy chunk of revenue and have pockets full of loose change. This is good because operating generative AI services at scale is astronomically expensive, especially when you’re opening them up to a class of users in consumers, where you’ll never be able to charge for the service. The capital expenditure (capex, as it’s referred to) required to stand up these operations is beyond almost any organization’s ability. As I mentioned beforehand, that’s just the way it suits those incumbents. 

What precisely the burn rate will likely be for operating large-scale generative AI is unclear to those of us outside the respective organizations. Still, those “cash on hand” figures will be interesting to watch if you’re interested in keeping a rough track. They’re a close-to-real-time tracker on what the long-game bet to own the AI market might be costing.

Robot image via Microsoft Copilot Designer

Leave a Comment

We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.

Work Intelligence Market Analysis 2024-2029