Everything To Everyone

Recently, word emerged that Amazon had topped Walmart in quarterly revenue for the first time, a big deal and a surprising one. I was sure they had done so five years ago.

The reason the company was able to do so is in large part a reflection of the game they’re playing, and the fact that people consider easy access to goods more desirable than cheap access.

But while Amazon has long been considered The Everything Store, we don’t necessarily want to rely on them for everything.

A fascinating document of when Amazon flops hard recently emerged on LinkedIn, with Ethan Evans, the retired vice president of Amazon’s Prime Gaming, reflecting on the company’s inability to compete in the space effectively. After all, people don’t tend to think of Amazon as a place to buy digital games. They buy full titles from Steam, or if they’re not choosy, subscribe to Xbox Game Pass.

Amazon isn’t really part of that discussion, despite owning a key part of the gaming ecosystem, the livestreaming network Twitch. To hear it from Evans, the reason for that is essentially that the company failed to understand the value proposition that Valve in particular had created around Steam:

At Amazon, we assumed that size and visibility would be enough to attract customers, but we underestimated the power of existing user habits. We never validated our core assumptions before investing heavily in solutions. The truth is that gamers already had the solution to their problems, and they weren’t going to switch platforms just because a new one was available.

We needed to build something dramatically better, but we failed to do so. And we needed to validate our assumptions about our customers before starting to build. But we never really did that either.

Steam, beyond being a game store, is also sort of a cultural home for gaming. It often plays matchmaker between you and the game you’re playing. After all, if you’re buying a title, you’re likely committing a significant amount of time to it. Steam’s discovery capabilities are distinctly different than what Amazon does when you’re trying to buy a new book or some home office supplies. I’m sure folks will argue otherwise, but I think Steam is in many ways a better experience for buying games than physical media—because it explains what you’re getting into.

It of course goes deeper than that. Valve understands the culture of gaming, whereas Amazon understands how to build an effective bundle that people will pay close to $150 per year to access. They’re playing a different game, and they’re likely to get different results.

Amazon’s strategy for Prime works when you’re trying to sell commodity goods. But for games, we generally want more than just convenience from how they’re delivered, because we spend a lot more time with them than we do movies or even TV shows. A good game can last you years. Amazon is not built to deliver that. Steam is.

Apple’s Not-So Intelligence

Amazon’s failings with gaming I think help explain the mess Apple has created for itself with Apple Intelligence. The company decided to put most of its features behind a “when it’s ready” wall, despite the fact that it was promoting those features in advertising that attempted to sell the public on their latest models.

Now we’re a full six months past the announcement of Apple Intelligence’s features, and observers are getting antsy. This week, it came out that some of those observers are within the company’s own rounded walls. Apparently, according to reporting from Bloomberg, no less than Apple software chief Craig Federighi put his thumb down regarding the Apple software efforts, finding that in his own personal testing that they didn’t actually work as expected. Now, as John Gruber points out, they’re getting straight-up delayed.

The reason Apple is struggling, to anyone keeping an eye on this discussion for a while, is obvious. LLM-style artificial intelligence, like it or not, is a collaborative effort, one built on research, iterative releases, and information-sharing. Apple is a company that has tried to bring as much of its stack in-house as it possibly can, sharply working against this collaborative nature. And it favors the well-orchestrated launch over the iterative release. It’s trying to do something really hard here, and (while they can still fix it) the ongoing challenges hint that it may not be in their wheelhouse.

Seemingly, Apple’s every decision—to lock out collaborators like NVIDIA, Intel, and AMD in favor of its own chipset, to ignore open standards like OpenCL in favor of its own technologies, and so on—exists to meet its goal of vertical integration. Hell, it just released its own modem. But in the case of artificial intelligence, vertical integration isn’t as great a fit. For Apple, it has long been clear that user-facing AI is not a core competency, based on how bad Siri has been for literal years.

Apple could be doing more to lean into this use case over whatever it’s trying to do with Apple Intelligence.

So, in this particular case, its locked down nature is biting them in the ass. Ironically, the one area where Apple’s AI game looks strong is with its desktop Macs, where local LLMs are showing potential. The reason? Apple Silicon’s unified memory allows the GPU to eat up more RAM than you can find on an NVIDIA card.

Apple has found ways to enter new markets in the past without losing their DNA. The iPod is an excellent historic example. But that’s because it played to the company’s cultural strenghths. Apple Intelligence does not.

Ultimately, I don’t think people use iPhones just to get access to AI. Problem is, Apple bet big on it, and changing course would make the AirPower flop from 2019 seem like a modest blip.

At some point it might just save more face to hand this functionality off to a partner that is better suited for this work. (Jensen Huang is waiting for your apology call, Tim.) It’s OK for Apple not to be in every pocket of our technology lives.

Too Big To Focus

The thread connecting companies like Apple and Amazon in this case is, simply put, one where these companies are so big that they must have an answer to literally every technology trend that passes through the ether.

Other companies are better suited to this moment. We bitch and moan about Google breaking its search—me, enough that I created a widely used workaround for it. But the fact is, Google started with an algorithm. The work that led to AI is in the company’s DNA in a way that famous flops like Stadia and Google+ were not.

Apple, on the other hand, has a close-to-its-chest culture that works against the very collaborative, data-soaked nature of artificial intelligence research, and the ramifications of those initial failings seem to be catching up with the company. This is a company that, when presented with the fact that high-end digital graphics teams were going all-in on CUDA, it let an old grudge get in the way of working with an essential partner, even at the cost of support from Hollywood. The only reason Apple is even in the 3D or LLM discussion now is because its GPUs do something that NVIDIA’s cannot.

Given my headline/subject line, someone would probably yell at me if I didn’t include this utterly fitting song for this story, so here it is. Still holds up. Why wasn’t Everclear a bigger band?

In the case of Amazon Prime Gaming, Ethan Evans at least tried to find a path forward from that failure. His post linked to a guest item in his newsletter where the author had this to say: “Talk to Real Customers Before Writing Code.” I think in the cases of these massive companies trying and failing to compete in these spaces far removed from their core competencies, too many leaders are saying “we need a response to this,” and not enough people in the room are responding, “but do we really?”

After all, it’s not like Apple’s going to lose the iPhone market overnight just because Siri sucks. Siri has sucked for a long time, and it’s had no real effect on their bottom line.