Meta Platforms (META)

The sleeper story is MSL has become Meta's Manhattan Project - six months in, they're so capacity constrained that they're restructuring their entire capital allocation and hinting at positive net debt to fund what Mark called 'the highest talent density lab in the industry.' The obvious headline is revenue acceleration, but strip out the 4-point FX tailwind and focus on the real story: 30% engineer productivity gains from AI coding tools are letting them do more with less people while spending $115-135B on compute infrastructure. The binding constraint isn't demand (they wish they had more capacity for ROI-positive core business uses) - it's talent and energy/power for the data centers. Most revealing quote: Mark's 'answers may be somewhat unfulfilling' on MSL progress is a classic Zuckerberg confidence downgrade from Q3's aggressive timeline promises. My view changes if they give specific MSL model capabilities and launch dates next quarter - the infrastructure spend only makes sense if the AI products deliver, and right now we're taking that on faith.