RSS Amplifier

Tech Support with Dan Ackerman · Jun 9, 2026

The Biggest Sign of AI Blowback

Read inOriginalenesdefrptjaittr
0
Sign in to vote or save

Dan Ackerman · Tech Support with Dan Ackerman

GitHub, often the canary in the coal mine for so much in tech, is shifting its AI services billing away from a flat subscription. “The company is switching its billing system from a flat subscription rate to a token-usage system that has the potential to bill users at a significantly higher rate,” TechCrunch reports.

I’ve written before about how the end of these generously subsidized rates may lead to AI sticker shock, especially as so many companies are engaged in performative AI tokenmaxxing exercises that literally challenge employees to burn as many tokens as possible.

Now, execs are abruptly backpedaling, working to “rein in freewheeling AI use,” says the Wall Street Journal: “Most said they don’t expect dramatic cuts to AI spending, given that many CEOs still have a mandate from Wall Street to show off how AI-forward their organizations are. Instead, companies will find uses for cheaper models and steer most employees to low-cost workflows, reserving the high token use for software engineering teams.”

For GitHub Copilot users, Ars Technica says many are “reporting some extreme sticker shock as they realize just how quickly their previous ‘normal’ usage is burning through their newly limited monthly allotment of AI credits.”

The free lunch era of AI may be ending, but the pack-your-own-lunch era is just kicking off.

From my perch on the hardware side of things, I can only imagine this will lead to more individuals and companies looking to run their own AI on their own hardware, rather than use pay-per-token cloud-based frontier models that live in massive data centers.

(We dug into the best local models for AI at MC News, where we recently looked at Gemma vs Qwen and did a deep dive into Qwen 3.5.)

To that end, you’ve also got high-end hardware like the Nvidia DGX Spark and the upcoming AMD Ryzen AI Halo, both able to run massive local models—although any sufficiently powerful desktop or laptop can also run versions of new LLMs like Gemma 4 or Qwen 3.6. I also think you’ll see more and more PCs charging a premium price for up to 128GB of shared system memory (most laptops have between 8GB and 32GB of RAM), currently considered the gold standard for running local AI on a laptop.

Moving users to local AI software rather than a cloud service may be as easy as leaning into subscription fatigue, leveraging the burnout from endless monthly fees for streaming, apps and games. How many AI tools are you willing to subscribe to? One or two? A half-dozen? That’s going to be harder math to juggle when you can swap between Gemma, DeepSeek, Qwen, Llama, etc. (and multiple weights and versions of each) at will, with no up-front costs beyond the computer hardware you’re running.

The free lunch era of AI may be ending, but the pack-your-own-lunch era is just kicking off.

If you want a way to quantify the success of Apple’s MacBook Neo, look no further than the tremendous pressure it’s put on PC makers to create reasonably priced laptops that look good and perform well. After 20 years of reviewing laptops, I can tell you that delivering most of a MacBook Air’s advantages for $599 is a massive wake-up call for the Windows space.

For example, Qualcomm’s new Snapdragon C chip is “taking aim at the MacBook Neo with a new System-on-a-Chip (SoC)...designed for low-cost devices that are expected to start at $300 later this year,” Windows Central reported recently.

And with Google essentially replacing Chromebooks with what it calls the Googlebook, it opens up the budget laptop battleground to many new competitors.

I have no doubt we’ll see more high-design laptops, decently spec’d and reasonably priced, going into the holiday 2026 season (with a few sneaking in for back-to-school season). Nvidia’s new RTX Spark SOC could eventually play a role there, too.

Ironically, while we may be getting more new, high-quality $599 laptops, so many other existing products are seeing price increases, even for hardware that’s years old.

Valve’s Steam Deck, which was as low as $399, has seen the prices of its two current models jump by as much as $300. The 512GB OLED model has went from $549 to $789, and the 1TB version, from $649 to $949. Keep in mind this is hardware that was most recently updated in 2023, with the core components virtually the same since its 2022 debut.

We’ve seen this with other gaming platforms as well. The base PS5 jumped from $499 to $599 (or $649 for the disc version), while the PS5 Pro went from $749 to $899. The Xbox Series X launched at $499 in 2020 and is now $649. The Switch 2 from Nintendo is scheduled to go from $449 to $499 later in 2026.

It’s this push and pull between two competing price pressures—increases driven by supply issues around RAM, storage and GPUs; and decreases driven by Apple’s MacBook Neo and its emerging competitors—that make shopping for tech feel like an especially confusing experience right now.

I tried to avoid the hype train, but I had to play 007 First Light, and it’s (almost) as good as everyone says, combining elements from IO’s own Hitman series with the classic third-person action that feels a bit like the Uncharted series.

Current read: A pre-release galley of “Monsters of Ohio“ by the great John Scalzi. Reminds me a bit of Widow’s Bay on Apple TV -- spooky, with stabs of dry humor. Keep an eye out for it later this year.

And if you missed it last time, here’s my longform video on the secret origin of the very idea of robots and AI. It all started with the 1920 play R.U.R., which coined the term “robot” and introduced AI-powered replicants (à la Blade Runner) that reshape the world’s economy. It’s a story I’ve been obsessed with for years, and the lessons of R.U.R. seem to come into sharper focus with every AI or robotics advancement.

Read the original on danackerman.substack.com ↗

Comments

Nothing yet. Say the first thing.

    Sign in to join the conversation.