Apple Just Raised MacBook Prices Over the 2026 Memory Crisis. Should YC F26 Founders Still Build Hardware or On-Device AI?

On June 25, 2026, Reuters reported that Apple is raising the prices of MacBooks and iPads, and the reason was not tariffs or a new chip. It was memory. The story shot to the front page of Hacker News with nearly 800 points the same week, because every builder reading it understood the second-order effect: if the most powerful hardware company on earth can't absorb DRAM costs, neither can your seed-stage startup.

If you're applying to YC F26 with anything that touches hardware, robotics, or on-device AI, a partner is going to wonder whether the 2026 memory crisis quietly breaks your unit economics. Here's how to think about it, and what to put in your application.

Why did Apple just raise prices, and how bad is the memory shortage?

The short version: AI ate the memory market. Hyperscalers buying High Bandwidth Memory (HBM) for GPU clusters have pulled production capacity away from ordinary DRAM. Micron has described a roughly 3-to-1 conversion ratio, meaning every bit of HBM capacity it adds costs it about three bits of conventional memory. AI workloads are projected to absorb around 20% of global DRAM wafer output in 2026, leaving commodity memory structurally undersupplied.

The price moves are not subtle. DRAM prices surged on the order of 80% to 100%-plus year-over-year heading into 2026, with Gartner forecasting a memory cost surge well above prior cycles. A 32GB DDR5 kit that sold for under $90 in early 2025 has been averaging north of $500. Framework raised DDR5 upgrade prices by about 50% on its laptops. Micron even retired its consumer-facing Crucial brand in February 2026 to focus on enterprise AI buyers. When Apple passes costs to customers, that's the macro reaching the most efficient supply chain in the world.

What does a DRAM shortage have to do with my YC application?

YC funds startups partly on whether the business can actually work at scale. A memory shock shows up in three places a partner cares about: your bill of materials, your gross margin, and your dependence on a commodity you don't control. If you're shipping a physical device, your single most expensive component may have doubled since you wrote your first projections. If you're running inference, your cloud memory and GPU costs are riding the same wave.

The mistake is pretending it isn't happening. The founders who stand out right now are the ones who name the constraint and show they've designed around it. YC partners have seen enough cycles to know that the best hardware companies, going back to the ones Paul Graham wrote about in 'The Hardware Renaissance,' treated component economics as a core part of the pitch, not a footnote.

Is it a bad time to apply to YC F26 with a hardware startup?

No, but the bar moved. Y Combinator has been visibly funding more hardware and deep tech. Zettascale (YC S24) is building AI chips and hiring chip engineers; the batch lists are full of robotics and physical-world companies. YC is not retreating from atoms. What's changed is that 'we'll figure out cost at volume' is no longer a credible line when the cost curve is moving against you in real time.

If you're building on-device or edge AI

This is the place where the memory crisis is actually a wedge, not just a risk. Cactus (YC S25) made the front page of Hacker News by distilling a model down to 26M parameters that runs locally. The whole premise of on-device AI, smaller models, less memory, no per-query cloud bill, becomes more valuable, not less, when memory is scarce and expensive. If your startup makes AI run in less RAM, the 2026 shortage is the strongest possible tailwind, and your application should say so directly: 'Memory just got 2x more expensive, and we need 90% less of it than the cloud incumbents.'

If you're building hardware or robotics

Show your bill of materials with current 2026 memory pricing, not 2024 pricing. Explain your design choices: are you using less memory by design, locking in supply, or building margin that survives another price doubling? A founder who walks into the interview and says 'here's our BOM at today's DRAM spot price and here's why we're still at 60% gross margin' instantly reads as someone who has done the work.

How should you address the memory crisis in your YC F26 application?

Write one or two specific sentences, not a disclaimer. Name the constraint, quantify it, and show your response. Something like: 'DDR5 prices roughly doubled in the past year as AI demand absorbed DRAM supply. Our device uses 4GB instead of the typical 16GB, so the shock raised our BOM by only $6, and we've secured pricing with our supplier through Q2 2027.' That's three sentences that tell a YC partner you understand your own business better than they do.

This is exactly the kind of claim worth pressure-testing before you hit submit. Reading your own application, it's easy to assume your memory math is obviously convincing when a reviewer will read it as hand-waving. Getting brutally honest feedback from people who've actually sat in the YC interview room, which is what we do at YC Roaster, is the fastest way to find the soft spots before a partner does. A founder who has rehearsed the 'what happens if memory doubles again' question answers it in ten seconds instead of freezing.

The one-line version

The 2026 memory crisis isn't a reason to skip YC F26. It's a filter. Founders who ignore it look naive; founders who quantify it and design around it look like operators, and the on-device AI teams who turn scarce memory into their core advantage may have just been handed the best market timing of the batch. Put the number in your application, and be ready to defend it.