#388: Is Microsoft Really Losing Money On GitHub Copilot?, & More

To their delight, developers are coding twice as fast[1] these days, thanks to Microsoft’s GitHub Copilot, but some observers are wondering about its economic viability. Last week, The Wall Street Journal (WSJ) claimed[2] that Microsoft is losing $20 a month on average per user, each of whom pays $10 per month for Copilot.

In response, Microsoft’s VP of Product, Mario Rodriguez, declared[3] that Copilot is thriving, generating revenue of $100 million at an annual rate and not losing money. Launched in October 2021, had it been a standalone company, GitHub Copilot would have enjoyed one of the fastest ramps in the history of software-as-a-service (SaaS). Our research—pace WSJ—highlights AI cost declines of 70% per year[4] as well as continuous price drops[5] on large language model application programing interfaces (APIs).

Microsoft plans to integrate additional Copilots into its broader Office 365 suite at a monthly price of $30 per user on November 1. While both Microsoft and Google seem to believe that users will pay hefty monthly fees for AI capabilities, other companies like Zoom[6] and Replit,[7] hoping to broaden their reach, are rolling out AI features for free.

While robotics has evolved through specialization—washing machines, dishwashers, spot welding robots, robot vacuums—the convergence of artificial intelligence, computer vision, and robotics could be opening the door to general purpose robotics platforms (GPRP). Adding traction to GPRPs is a handful of companies creating humanoid robots

In our view, generalized applications need not be confined to human-like forms. The evolution of calculators and computers is instructive. Early calculators and computers seemed similar, but their respective design trajectories diverged considerably. Generalizable computers became the much larger opportunity. While many specialty robotics companies are likely to find success over time, generalizable robotics could become the platform for several “killer apps” that take the market.

In hopes of enhancing cognitive treatments and capabilities for neurological disorders, Elon Musk’s “Neuralink” is developing brain-machine interfaces (BMIs) that link the human brain with computers. Building on the work of other technologies like deep brain stimulation (DBS), electrocorticography, and machine learning, Neuralink provides high-resolution, low-latency monitoring of brain activity. Implanting into a patient’s sensorimotor cortex a previously unthinkable 2000+ electrodes distributed across hair-like threads, Neuralink has significant potential to transform medicine and acclerate the link between humans and artificial intelligence.

Neuralink’s “N1” module—and the “R1” surgical robot required to implant it—could help amputees achieve more natural and fluid motion as their own thoughts control prosthetic limbs. The technology also could bypass spinal cord injuries and transmit brain signals directly to devices that help amputees regain mobility and control. Longer term, BMIs could treat amytrophic lateral sclerosis (ALS), Parkinson’s disease, epilepsy, sensory disorders, stroke-related brain injury, and many other conditions. Indeed, Neuralink already is enrolling a clinical trial for ALS patients.

Assuming FDA approval in ALS and cervical spinal cord injuries and a modest rate of adoption by 2025, our preliminary research suggests that annual revenue just for N1 devices could surpass $220 million by 2030.

[1] Kalliamvakou, E. 2022. “Research: quantifying GitHub Copilot’s impact on developer productivity and happiness.”
[2] Dotan, T. 2023. “Big Tech Struggles To Turn AI Hype Into Profits.” The Wall Street Journal.
[3] “AI Engineer Summit 2023 — DAY 2 Livestream – YouTube.”
[4] ARK Investment Management LLC. 2023. “Big Ideas 2023.”
[5] The Cognitive Revolution. 2023. “OpenAI Price Drop: [Another] Megathread.”
[6] Zoom. 2023. “Getting started with Zoom AI Companion features.”
[7] Replit AI Team. 2023. “Announcing Replit AI for All.”