It’s 2026, and I’m starting to think the console wars were never really about consoles.
I’m in my mid-thirties, watching gaming built around hardware and platforms. Some called it the console wars. Some called it competition.
Either way, I’ve grown less fond of buying hardware for an exclusive these days.
Look at the supply chains. Look at the economic status of the players. It’s clear the next-gen platforms are going to lean on a few SKUs, with more frequent refresh cycles. The SoC model is the future – of scaling, of efficiency, of optimisation. One tightly integrated chip, iterated often. That’s the direction of travel, and nobody’s pretending otherwise anymore.
And there’s a reason the whole industry is converging on it at once. When supply chains are tight – and they’ve been tight since 2020 – you don’t win by offering infinite variety. You win by consolidating a handful of designs that scale. Look at who’s actually building the silicon: AMD is making the custom SoC for the next Xbox and the custom chip rumoured for the PlayStation 6, both pointed at 2027. One chip company, two “rival” consoles. That isn’t a coincidence. That’s what constrained supply looks like – a few large players consolidating designs that scale, and everyone else renting capacity from them.
Windows could never really get the formula right
Windows made the opposite bet. Infinite customisation. Any CPU, any GPU, any amount of RAM, any of a dozen driver versions, sitting on an OS that updates constantly. Freedom – and a combinatorial nightmare underneath it.
Put a number on it. When NVIDIA ships a single “Game Ready” driver, it validates that one driver across more than 4,500 combinations of GPU, CPU, RAM and OS, covering 89 different GPUs going back more than a decade. That’s one vendor’s slice of the matrix. A studio shipping a PC game is firing into a space where a rendering bug can show up on one driver version, on one card, for a sliver of players – and never reproduce on the machine the developer is sitting at. Drivers update monthly. Every update can quietly break something that shipped fine last week.
That’s the tax on customisation. Every degree of freedom you hand the consumer multiplies the test space the developer has to cover, and the surface area where bugs hide.
This is exactly what consoles were built to kill. A console exposes a single hardware configuration and an API that sits lower than any desktop driver. Build once, to a fixed target, and it runs the same on every unit in every living room. Fixed hardware, fixed economics, a known quantity for developers. The “limitation” everyone mocked – you can’t upgrade a console – was the whole point. Predictability is a feature, not a compromise.
Console makers have been on the SoC model for years for the same reason. One chip. Known thermals. Known performance. Iterate the design, don’t fragment it.
Now the holdout is converging. Microsoft has confirmed its next-generation Xbox – codename Project Helix – will be powered by a custom AMD system-on-chip, co-designed for the next generation of DirectX and FSR, with developer alpha kits slated to ship in 2027. The kicker: it’s built to run full Windows 11, with the console interface delivered through a “full-screen experience” layered over the Xbox PC app. Sarah Bond has been blunt about the ambition – an Xbox experience “not locked to a single store or tied to one device.” Anything can be an Xbox.
Which is another way of admitting Windows never got the formula right on its own. For thirty years it was the open, fragmented platform that couldn’t decide if it was a games machine or a spreadsheet. Now it’s borrowing the console playbook – custom silicon, one coherent stack – to fix a problem it created.
And somewhere in between all of this, Mac arrived. Hardware that quietly became a platform – for gaming, for AI, for everyone else. Competition.
The yield is the strategy
Here’s the part that gets left off the keynote slides. The reason a few designs beat infinite variety isn’t taste. It’s physics, and it’s math.
Chips are printed on silicon wafers, and wafers have defects. The simplest model of how many usable chips you get – the yield – is the Poisson model:
Y = e^(−A·D)
Yield falls exponentially as the chip area A grows, for a given defect density D. Double the die size and you don’t shave a little off the yield, you fall off a cliff. The slightly more realistic Murphy model – Y = ((1 − e^(−A·D)) / (A·D))² – bends the curve more gently but tells the same story. Either way the lesson is identical: smaller, repeated, well-characterised designs yield better than big bespoke ones. Fewer SKUs, made over and over, is how you get high-quality output out of an absurdly expensive fab.
And here’s the elegant part. When chips come off the line at different performance levels – or with a flaw in one corner – you don’t bin them in the bad sense and throw them out. You bin them in the good sense and sort them. Chip binning: the same wafer produces dies graded by how many cores work, what clock they hold stable, what voltage and heat they run at. The best ones become the premium SKU. The slightly-flawed ones become the cheaper SKU with a core fused off or a lower clock. One design, one production line, a whole product ladder out the other end – built from the natural distribution of what the fab actually produced.
So when I say “a few SKUs with frequent refresh cycles,” that isn’t a marketing instinct. It’s the yield curve talking. Consolidate the designs, run them hard, bin the output into a ladder, iterate next cycle. It scales because the manufacturing scales.
Windows-style fragmentation is the anti-pattern to all of this. Every config you support is another bin to validate, another row in that 4,500-line test matrix, another place a defect can hide – in software this time instead of silicon. Same math, one layer up.
The Corolla isn’t boring
People say a Corolla is boring. They say Apple iterates its designs too slowly. Same complaint, and they’ve got it backwards.
Toyota’s reliability comes out of the Toyota Production System and the philosophy underneath it – kaizen, continuous incremental improvement. Not dramatic overhauls. Small, constant, compounding refinements, with every worker empowered to stop the line the instant a defect appears (jidoka) so a bad part never flows downstream. Toyota takes in over a million employee improvement suggestions a year and implements the overwhelming majority of them. The result: J.D. Power consistently ranks it near the top for dependability, and the brand is a byword for durability, low running cost – and strong resale value. The “boring” is the point. Boring is what high yield feels like from the driver’s seat.
Apple runs the same philosophy, just on silicon and glass instead of steel. The industrial design changes slowly because changing it slowly is how you hold yields high and defects low across hundreds of millions of units. Slow iteration isn’t timidity. It’s the same disciplined incrementalism that lets a Corolla outlast cars that cost three times as much.
Put the three together and it’s one idea wearing three outfits. A bug in a PC game, a fault in a car, a dead die on a wafer – they’re all defects escaping into a space too varied to fully test. The fix is always the same: shrink the variety, standardise the configuration, improve in small steps, and let predictability do the heavy lifting. Faults in cars, bugs in fragmented PCs, scrap on a wafer – predictable configurations beat customisable ones on every single one.
Apple’s quiet position
Here’s the difference. While everyone else carried fragmented structure and fragmented distribution, Apple had the first sale. The long life expectancy – like a Corolla. Cross-integration with its own supply chains and product lines, top to bottom.
The numbers back the metaphor. Apple’s active installed base crossed 2.5 billion devices in early 2026, up from 2.35 billion a year before – roughly 150 million net new devices a year, every year. Services hit another all-time revenue record, up 14% year over year. That’s the machine: hardware that lasts, sells once, and then quietly compounds into a recurring services annuity. The Mac alone is doing north of $8 billion a quarter without anyone calling it a gaming brand.
It’s quietly spectacular when you sit with it. The first sale funds the device, the device feeds the services, the services raise the value of the next device. A flywheel most of the industry can only describe in a pitch deck.
The only thing Apple hasn’t solved for is the developer community. That’s it. That’s the whole gap.
The five-year device, the ten-year market
Here’s the design problem nobody states out loud, but everyone in this race is quietly trying to solve:
How do you build a device that’s good for five years for the average buyer – durable enough to live a second and third life on the used market for up to ten – while also keeping the high-income buyer who wants to upgrade every year, trades the old one in, and never really feels the depreciation?
Those two customers sound opposed. They’re not. They’re two ends of the same flywheel, and the bridge between them is resale value.
The data is stark. iPhones hold roughly 60 – 70% of their value after two years and around half after four; flagship Androids are down near 20% over the same window (BankMyCell, Swappie). Apple gives 5 – 6 years of iOS updates, so a four-year-old phone is still current, secure and desirable – which is exactly why it still commands a price. And the kicker: by some estimates close to half of new iPhone buyers now fund the upgrade by trading in or selling the old one. The used device becomes the down-payment on the new one.
Follow the loop:
slow, high-yield design
→ reliability + long software support
→ device still good years later
→ strong resale value
→ deep, trusted second- and third-hand market
→ high earner upgrades often (trade-in is cheap and easy)
→ their year-old device becomes the value buyer’s entry point
→ value buyer enters the ecosystem and starts paying for services
→ (loop)
The frequent upgrader and the ten-year hand-me-down buyer aren’t in tension. They’re feeding each other, and Apple monetises every turn of the wheel – the first sale, the services, the trade-in margin, the refurbished resale, and the new first sale the trade-in just financed.
That’s the formula behind the formula. Build for predictable quality, and resale value does the rest. It’s the Corolla strategy, applied to a computer you keep in your pocket – and the same logic that makes a “boring,” slow-iterating, hard-to-fragment device the most valuable thing in the room over a ten-year horizon.
The crack in the wall
Age of Empires just launched on Mac.
Let that sit for a second. The Definitive Edition arrived on macOS at the end of May 2026 – Apple Silicon native, M1 or newer, built with Feral Interactive and Microsoft’s World’s Edge. It’s the first major Microsoft-published title to land on the Mac since Psychonauts 2 back in 2021. A five-year gap, closed. And it didn’t arrive alone – it landed into a Mac that now has the Game Porting Toolkit, Metal, a dedicated gaming mode, and a growing shelf that already includes Cyberpunk 2077, Assassin’s Creed Shadows, and Resident Evil 4.
My nostalgia for this stuff is at an all-time high. And it’s peaking at the exact moment the Windows platform I grew up on has isolated itself in hardware cohesion so hard that it’s difficult to want to go back. Unless you’re on a handheld. Or waiting on the future of an open platform, if the available information is to be believed.
That’s the crack in Apple’s only wall. One AAA strategy classic at a time, the developer gap is starting to close – and it’s Microsoft, of all companies, helping close it.
The store, not the box
Here’s the part the hardware story keeps missing. The wall was never really about chips or ports. It’s about who owns the storefront.
Watch what happened when Apple tried to own it. They put Assassin’s Creed Mirage on stage at an iPhone launch, handed Ubisoft the full marketing machine – console-grade gaming in your pocket – and fewer than 3,000 people bought the full game. Roughly a 2.4% uptake on 123,000 downloads. Death Stranding and the Resident Evil ports did the same thing. Then Ubisoft stopped updating Mirage and it started crashing on launch. So the curated first-party showcase didn’t just sell badly – it left the people who paid holding a broken app. That’s the model Apple wanted developers to trust. Horrible.
Now watch the storefront route. Buy Cyberpunk on GOG years ago, and when it comes to Mac it’s just there – free, because you already own it, your save waiting in the cloud, cross-progression straight off your PC. It shipped on GOG, Steam, Epic and the App Store all at once. And the tell is in the file size: the GOG and Steam builds are 92GB, the App Store build is 149GB, because Apple’s rules force everything into a single bundle. The platform owner’s own store is the worse way to deliver the same game.
So the pattern is clean. The store tied to the hardware loses. The store that abstracts the hardware wins.
That’s why AoE2 matters more than the nostalgia. It went Steam-first, with the Mac App Store version trailing later in the year. Feral spent a decade as the dependent porting shop, living off whatever the Mac and Linux markets threw them – and even they’re routing through Steam now. AoE2 isn’t proof that Apple cracked developers. It’s proof that Steam did. The uptake on Steam-for-Mac is the number to watch, because it’ll show the real potential on a platform where the first-party-store approach failed outright.
And it folds straight back into Helix. If the next Xbox is Windows underneath, Steam and GOG run on it natively – and the Xbox store becomes just another launcher on a box that was already a PC. Meet halfway on the hardware and you hand distribution to whoever owns the agnostic storefront. For most developers, that’s where the real clout sits. Not the box. The shelf.
One catch, said out loud: “storefronts win” mostly means Valve wins. Trade a console gatekeeper for a 30%-cut gatekeeper with even more reach and you haven’t escaped the toll – you’ve just changed who collects it. GOG and Epic are the only real pressure valves. And the Mac only works as the proof case because it’s open enough to run Steam at all. iOS still isn’t, which is the whole reason the App Store push had nowhere else to route.
The regulators are forcing the doors open
There’s a third force in the room now, and it’s pushing in Apple’s structural direction whether Apple likes it or not: the EU.
On distribution, the Digital Markets Act has pried the App Store open. In the EU, Apple now has to allow sideloading, third-party app marketplaces, and alternative payment systems – and when it dragged its feet on “steering” users toward cheaper options, the Commission fined it €500 million, with penalties of up to 10% of global revenue sitting behind the threat. Whatever you make of the security trade-offs Apple keeps warning about, the direction is unmistakable: the platform owner’s lock on its own store is being legislated away. Which is the same conclusion the storefront argument reached from the market side – the agnostic store wins, and now the law is helping it along.
On hardware, the consumer-rights push runs straight into the longevity flywheel. The EU already mandates USB-C, at least five years of software updates, and seven years of spare-part availability. The Right to Repair Directive and Regulation 2023/1542 go further – from February 2027, batteries have to be user- or professionally replaceable, and parts-pairing (software locking a device to its original components) is banned. As one legal write-up put it, Brussels has shifted the default assumption of how long a phone should last, and parked the responsibility for that on the manufacturer.
Read that against the five-year-device thesis and it’s almost funny. Regulators are forcing exactly the conditions that make the resale flywheel spin harder: longer support, easier repair, longer usable life, a bigger and more trustworthy second-hand market. The company built around durable hardware and recurring services is the one least disrupted by being ordered to make durable hardware that lasts. The fragmented, hardware-churning model is the one with something to lose.
The investor read
So assess the companies, not the fandoms.
The market itself is maturing, not exploding. Newzoo put global gaming at roughly $197 billion in 2025, up about 7.5%, with 2026 forecast around $208 billion and the player base near 3.6 billion. Growth is real but slowing relative to the online population. The interesting line isn’t the total – it’s the mix: PC gaming grew over 10% to about $43 billion in 2025 and is projected to overtake console revenue within a year or two. The open, software-led side of the market is winning. The boxed-hardware side is bleeding.
And the bleeding shows up plainly in Microsoft’s own filings:
- Xbox hardware revenue fell 32% in the December 2025 quarter, the most lucrative quarter of the year, on lower console volume.
- Overall Xbox gaming revenue was down 9%, with content and services down 5% – the first decline in that line since FY2023.
- Zoom out and it’s a trend, not a blip: Xbox hardware was down 13% in FY24, then 25% in FY25.
- All of this after two console price hikes in 2025 and a 50% increase on the top Game Pass tier.
Apple, over the same window, kept printing services records on a still-expanding installed base. Two companies, two directions – and the difference traces straight back to the yield curve, the resale flywheel, and who actually owns the shelf.
So here’s the prediction.
Perhaps software is software, and the economics of gaming are simply about the platforms that can distribute to enough people to make financial sense. Distribution is the whole game now. Not the chip, not the exclusive – reach.
And perhaps Xbox – Project Helix, the Windows hybrid, “anything is an Xbox” – is purely a grab at being an all-rounder device. A play for Mac mini and Mac Studio volume, reach, and refresh cycles. An attempt to claw traction back into an ecosystem that struggles with identity outside of business volume, at a time when the consumer market has been shifting away from it for years. Microsoft watched Apple turn one hardware sale into a decade of services, watched it consolidate a few SKUs that scale, and now it wants the same flywheel – it just has to dismantle the console it built to get there.
Time will tell
Services have always been meant to be agnostic. Play anywhere, on anything, no allegiance required.
But stickiness is where the long-term revenue sits. The lock-in, the installed base, the device that lasts long enough to keep selling you things. That’s the part the agnostic story conveniently skips.
Whoever cracks that formula – predictable quality that scales, distribution wide enough to make sense, and stickiness deep enough to compound across a ten-year resale chain – is bound to win.
Right now, only one of these companies is being quiet about it.