In the 1980s, a Cambridge researcher named Tony Robinson was already applying neural networks to speech recognition. Decades later, the company he founded sits inside other firms’ captioning tools, contact-centre software and voice agents, often unseen by the end user.
This piece sets out how Speechmatics makes money, what it charges, who pays, and what the public record does and does not tell us about its finances.
Why Speechmatics Sells to Businesses, Not Consumers
Speechmatics builds speech-to-text software that other companies embed in their own products through an API, so its customers are businesses.
The company says founder Dr Tony Robinson pioneered applying neural networks to speech recognition at Cambridge University in the 1980s.
Companies House shows Cantab Research Limited, which Bloomberg lists as trading as Speechmatics, was incorporated in February 2006. Katy Wigdahl is chief executive, and Robinson remains on the board.
Customers can run the technology as a cloud API, on their own servers or on a device.
It supports more than 55 languages.
How Speechmatics Makes Money?
Speechmatics makes money by charging businesses for each hour of audio it transcribes, with pre-recorded rates from $0.12 an hour.
That usage-based logic sits at the heart of the Speechmatics business model.
A customer whose product handles more audio pays more, and one that handles less pays less.
New accounts receive $100 in free credits with no card required, so a developer can test the service before spending anything. Larger customers deal with Speechmatics’ sales team and receive dedicated support.
At its simplest, the company meters the work its software does.
Speechmatics Pricing: From $0.12 an Hour to $1,250 a Month
Speechmatics charges $0.12 to $0.38 an hour for pre-recorded transcription, depending on the model, and since 1 October 2026 offers pay-as-you-go, credit packs and monthly subscriptions.
The company’s 16 September 2026 announcement says the $1,250 monthly subscription carries a 15% discount, rising to 25% when a customer pays for twelve months upfront.
A $25 credit pack gives smaller teams a low-cost way to start. The old 20% discount for pay-as-you-go customers above 500 hours a month ended on 1 October, with discounts moving to credit packs and subscriptions.
Speechmatics also lists a 33% discount for customers who enable model training on their account.
The Add-Ons That Stack on Every Bill
Beyond core transcription, Speechmatics sells add-ons such as translation at $0.49 an hour, text-to-speech at $0.011 per 1,000 characters and a speech-to-text API built for voice agents.
Other per-hour add-ons include chapters (0.30), topics(0.15), summaries (0.09) and sentiment(0.09), according to the pricing page. These are a second layer, because each is billed on top of the transcription rate.
The first million text-to-speech characters are free.
Agent STT, powered by a model called Linden, launched on 17 September 2026 at a launch price of $0.30 an hour.
We found no published split showing how much each product line contributes.
Who Pays Speechmatics? The Brands Behind the Brands
Speechmatics’ customers include media, captioning, contact-centre and healthcare companies that build transcription into their own services.
Its June 2022 funding announcement named 3Play Media, Veritone, Deloitte UK and Vonage, alongside government departments. Its partner page lists Panopto, 3Play Media, Veritone, DeepL and Syncwords in media and captioning, and its case studies feature Humetrix and Edvak in healthcare and Liquid Voice in contact centres.
In August 2026 it said Norwegian conversational AI company boost.ai runs Speechmatics in production for its voice agents. Each of those partners embeds the API in a product of its own.
Speechmatics often sits out of sight.
In a September 2026 post about customer Tresic, it noted that end customers only ever see their own provider’s brand.
Speechmatics Revenue: What the Public Record Can’t Tell You
The most authoritative source for Speechmatics revenue is the group accounts Cantab Research Limited files at Companies House, the latest covering the year to 31 December 2024.
Third-party estimates vary. CB Insights lists $12.72 million for the 2022 financial year, Wikipedia gives €11.3 million for 2021, and Dealroom shows only a $10–50 million band. Periods, currencies and methods differ, so we treat these as indicators rather than audited figures.
We would rather send you to the filing than hand you a number we cannot stand behind. Read the accounts before you quote any revenue figure. Until then, how Speechmatics makes money is clearer than how much it makes.
Speechmatics Funding: Decades of Research, Then a $62M Series B
Speechmatics raised a £6.35 million Series A in October 2019 and a $62 million Series B in June 2022, led by AlbionVC and Susquehanna Growth Equity respectively.
The wider Speechmatics funding story starts earlier. The company’s About page says IQ Capital invested at the end of 2016 and Amadeus Capital Partners in late 2016, roughly a decade after incorporation. AlbionVC led the Series A, with IQ Capital and angel investors taking part.
AlbionVC and IQ Capital, both existing investors, took part in the Series B. Speechmatics said it would spend the money on US and Asia-Pacific expansion, data-centre capacity and research.
Susquehanna’s Jonathan Klahr said the firm began tracking the company after its portfolio companies reported that Speechmatics beat Big Tech options on accuracy.
Speechmatics Business Model: Where It’s Strong, Where It’s Exposed
Speechmatics’ own comparison pages name nine rivals, including Deepgram, AssemblyAI, Google, AWS, Microsoft Azure and OpenAI.
The Speechmatics business model has clear strengths.
Investors cite accuracy, and customers can deploy in the cloud, on-premises or on a device. Speechmatics says an independent Pipecat benchmark in August 2026 gave it the lowest pooled word error rate, 1.07%, among 12 services tested.
The risks are structural. Per-hour pricing ties income to customer volumes, and it competes with cloud giants and specialists.
We found no published customer count or revenue split, so we cannot judge from outside how far it depends on a few large customers.
Three Lessons for Founders From Speechmatics’ Long Road
The clearest lesson from Speechmatics is to sell the layer beneath other companies’ products, let customers start free, and keep refining how you charge.
The first point follows from the customer list above: partners build their products on its API. The second is visible in the $100 credit and the $25 pack. The third shows in the October 2026 pricing changes.
We think the timeline is the most instructive part.
Speechmatics funding arrived late in the story: Robinson’s research began in the 1980s, the company was incorporated in 2006, institutional money followed in late 2016, and the Series B came in 2022.
It asks founders for patience.
FAQs
1. How much does Speechmatics cost?
Speechmatics charges $0.12 to $0.38 an hour for pre-recorded transcription, depending on the model, and new accounts receive $100 in free credits. Add-ons such as translation cost extra per hour. Per-hour rates like these are how Speechmatics makes money, and they change, so check the pricing page before budgeting.
2. Who founded Speechmatics?
Dr Tony Robinson founded Speechmatics. The company says he pioneered applying neural networks to speech recognition at Cambridge University in the 1980s, and he remains on its board. Katy Wigdahl is chief executive.
3. Who are Speechmatics’ main competitors?
Speechmatics’ own comparison pages name AssemblyAI, Deepgram, Google, AWS, Microsoft Azure, OpenAI, ElevenLabs, Gladia and Soniox. CB Insights separately lists Deepgram, Fano Labs and AssemblyAI as its top competitors.
Also Read: Who Really Owns the UK’s Booming Voice AI Industry?
Sources: Speechmatics (pricing, About and partner pages, funding announcement of 28 June 2022, blog posts, accessed October 2026), Companies House, UKTN, Bloomberg, CB Insights, Dealroom and Wikipedia. Figures reflect the latest data available at the time of writing.
The EP+ Editorial Desk covers UK startups, founder stories, and venture capital. All editorial content is independently produced and human-reviewed before publication.
