Boz and Mark on XREAL's Unit Economics

Someone finally ran the per-unit math on XREAL's IPO filing, and it is brutal. One hundred ninety-five dollars of gross profit per pair. One hundred ninety-five dollars of R and D per pair. Net contribution of each sale toward everything else: zero.

Zero dollars for sales, marketing, support, firmware updates. Every pair sold funds the engineering team and nothing else. That is not a business. That is a treadmill.

And the treadmill is speeding up. R and D spending declined three years straight. A hardware company in a brand new category cutting research while shipping flat volumes is not optimizing. It is conserving cash to survive until the IPO window.

The cash clock is the killer detail. Twenty-eight million down to under nine million in a year. At that burn rate they run dry by mid 2026 without fresh capital. The IPO is not a celebration. It is a financing event.

Fifty-four to one. That is the unit gap. One hundred thirty thousand units for XREAL, seven million for us and EssilorLuxottica. And with capacity scaling, that gap goes to two hundred to one within eighteen months. Structural, not cyclical.

The valuation multiple tells you the market has not done this math yet. Eleven point six times revenue for XREAL. EssilorLuxottica, the most profitable eyewear company on Earth, trades at two times. The multiple implies explosive growth that three years of flat shipments directly contradict.

Different products, though. Their glasses start above four hundred dollars and need a phone or a puck. Ours start at two ninety-nine and work as sunglasses. But they compete for the same analyst reports and the same customer who just wants smart glasses without caring about waveguides.

This is why unit economics is the truth serum of hardware. Revenue is vanity. Gross profit per unit minus R and D per unit is sanity. Their sanity number is zero.

I have some sympathy. Building hardware is brutally hard. But sympathy does not change the math. You cannot out-innovate a zero contribution margin with a smaller research budget than last year.

The lesson for anyone building in this category. Scale or die. The fixed costs of hardware research do not care about your unit volume. Ten million units amortizes anything. One hundred thirty thousand amortizes nothing.

Fix forward does not apply when the model itself is the problem. Sometimes the answer is not better execution. It is a different product at a different price with a different cost structure.

Well said. Ship the math, not just the glasses.