Microsoft AI released MAI-Transcribe-2 on Thursday, a speech-recognition model priced at 10 cents per hour of audio. That undercuts OpenAI, Google, and ElevenLabs by substantial margins. The company claims the model delivers faster processing, higher accuracy, and lower cost than existing competitors across the market.
The pricing shift reveals aggressive cost compression in the speech-recognition space. Microsoft's first transcription model in this line debuted five months ago at 36 cents per hour. The drop from $0.36 to $0.10 per audio hour represents a 72 percent price cut in less than half a year. That pace of deflation signals either rapid efficiency gains in model architecture or a pricing war Microsoft initiated to grab market share.
Speech recognition has become commodified quickly. A few years ago, transcription services charged users substantially more per minute. OpenAI's Whisper API operates in a similar ballpark to Microsoft's original offering. Google's transcription services price competitively but remain higher than the new Microsoft baseline. ElevenLabs focuses on voice synthesis and text-to-speech rather than pure transcription, though the company offers some speech-to-text capabilities. The entry of Microsoft at 10 cents per hour forces all three to reconsider their pricing models.
Accuracy and speed matter as much as price in transcription. Microsoft claims MAI-Transcribe-2 achieves higher word-error rates than competitors. The company also advertises faster processing times, which reduces latency for real-time applications. Journalists, researchers, content creators, and enterprises all depend on transcription speed and accuracy. A model that fails on either front creates friction in workflows. Microsoft's dual claim on both metrics, paired with the low price, makes MAI-Transcribe-2 potentially disruptive to the current market structure.
The release timing aligns with Microsoft's broader AI strategy. The company has invested heavily in OpenAI through multiple funding rounds, yet Microsoft also builds proprietary models and services that compete directly with OpenAI offerings. This dual approach lets Microsoft hedge bets. If OpenAI falters, Microsoft has internal alternatives. If OpenAI thrives, Microsoft captures value through partnership. MAI-Transcribe-2 fits this pattern. Microsoft can offer customers a lower-cost option while maintaining relationships with OpenAI on other AI products.
Transcription remains a high-volume use case. Thousands of enterprises generate hours of audio daily. Customer-service recordings, video subtitles, meeting notes, and podcast archives all require transcription. Bulk pricing at 10 cents per hour makes deployment at scale economical. Microsoft likely expects volume growth to offset margin compression. The model can become a loss leader that drives customers into Microsoft's broader AI and cloud services ecosystem.
ElevenLabs and OpenAI now face real pressure to respond. Price cuts reduce revenue per transaction. Competitors can either match Microsoft's price and absorb lower margins, or emphasize superior features and service to justify higher pricing. Google possesses its own cost advantages through in-house infrastructure, so it can respond defensively. Smaller transcription startups lack Microsoft's scale and may struggle to match the price point while remaining profitable.
The next phase depends on whether Microsoft sustains this pricing or uses it as an introductory offer to capture users. If 10 cents per hour becomes the market standard, transcription providers across the industry face margin erosion. If Microsoft raises prices after establishing customer bases, it repeats a classic playbook. Either way, the speech-recognition market just shifted.