Episode
How Meta Is Using Its Own Chip to Cut AI Inference Costs
- Published
- Jul 9, 2026
- Duration seconds
- 618
- Processing state
not_requested
Actions
POST https://stenobird.com/v1/public/podcasts/the-ai-podcast-with-fexingo-artificial-intelligence-machine-learning-and-modern-ai-models-7872011/episodes/how-meta-is-using-its-own-chip-to-cut-ai-inference-costs/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/the-ai-podcast-with-fexingo-artificial-intelligence-machine-learning-and-modern-ai-models-7872011/how-meta-is-using-its-own-chip-to-cut-ai-inference-costs.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Meta just released Muse Spark 1.1, its latest AI coding model, but the bigger story is how the company is running inference on its custom Meta Training and Inference Accelerator chip. In this episode, Lucas and Luna break down the cost savings of owning your own inference silicon, why Broadcom's 11 percent weekly surge is connected to custom chip deals, and how the math of AI inference is reshaping the hardware landscape. They explain the difference between general-purpose GPUs from NVIDIA and AMD versus custom ASICs like Meta's MTIA, and what that means for the next wave of AI deployment at scale. If you follow AI hardware, this is the angle the headlines are missing. #Meta #MuseSpark #AICoding #CustomChip #MTIA #AIInference #Broadcom #NVIDIA #ASIC #Hardware #Technology #TechPodcast #ArtificialIntelligence #ChipDesign #InferenceCosts #FexingoBusiness #BusinessPodcast #AIPodcast Keep every episode free: buymeacoffee.com/fexingo