For anyone following the software industry, there’s been a little bit of snark about C3.ai (“C3”) over the years. Here’s a company that was founded by Silicon Valley royalty (Tom Siebel, who sold Siebel Systems to Oracle in 2006 for just shy of $6B), with seemingly limitless access to capital, that somehow seemed to be pivoting every few years to something new – from energy at first, to the Internet of Things, to Artificial Intelligence.
C3 also largely eschewed the startup echochamber – funded personally by its founder at first, it didn’t raise money from the usual VC suspects, target well-know startups as its first customers, or open source any AI frameworks, working instead with a small group of Fortune 1000 and government customers. As a result, it didn’t build the kind of buzz that often precedes the most notable startups on their way to becoming public.
Lo and behold, what emerges in this IPO is a solid company by enterprise software IPO standards, with $157m in revenue, growing 71% yoy, a 75% gross margin and a manageable $69m loss.
It will be interesting to see how the market reacts to this IPO.
On the one hand, C3 is not growing as explosively as a Snowflake, and there are a few concerns, around account concentration for example.
On the other hand, C3 is very clearly positioning itself as one of the very first enterprise AI companies to go public: its ticker symbol on the NYSE will be “AI”, and the term “machine learning” is mentioned 56 times in the S-1.
The tailwinds around the deployment of ML/AI in the enterprise are very strong, and this IPO will be an interesting test for the continued appetite of financial markets for all things AI.
Here’s a tour of the S-1 and main characteristics of the business, that my FirstMark colleague John Wu and I put together.
- Founded in 2009, the company has taken a long road to scalable revenue, breaking the $100M recurring revenue mark only earlier this year. At 11 years since founding, C3 lands just above the median age for recent SaaS IPOs, but ranks in the lower half in terms of revenue scale, with above average growth rates.
- C3 hit its stride in FY20, growing annual revenue past $150M, and bringing in several larger contracts. Prior to FY19, revenue was largely concentrated in utilities and manufacturing, and at their assumed customer sizes, likely among a very small number of paying customers.
- C3 works has closed a number of large marquee logos, including Shell, AstraZeneca, conEdison, Koch Industries, Raytheon, and the US Airforce.
- C3’s revenue is highly concentrated both in terms of customers and sectors.
- The top two customers accounted for 26% (12% & 14%) of revenue in FY19, growing to 36% (26% & 10%) in FY20.
- As of July 31, 2020, the top four customers accounted for a whopping 69% of accounts receivable, and the top 3 customers represented 67% of accounts receivable in April.
- A major revenue driver for C3.ai is its strategic partnership with Baker Hughes, the oil and gas services company (both a customer and a partner). BH represented $39.5M of their total $157M revenue (25% of total) in FY20, and was on the hook for additional $11.5M as a partner reseller. The contract was downsized and modified to $27.2M in fees and $26.1M in resell for 2021, but that still represents a substantial portion of their revenue.
- Some interesting vanity metrics: 1.1B predictions a day over 4.8M production ML models, drawing data from 622M unique sensors.
In a way that somewhat echoes Palantir (see our Palantir S-1 teardown), C3 provides a “full stack”, bespoke platform for large enterprise customers.
It provides a “SaaS applications that enable the rapid deployment of enterprise-scale AI applications of extraordinary scale and complexity”, covering a wide range of commercial use cases, including energy management, predictive maintenance, AML, and predictive CRM.
C3 targets the type of customers that will want to outsource a large portion of their AI stack construction.
C3 views its full-stack solution as reducing complexity, leading to faster delivery, less maintenance, and faster time to value. Stitching together a comprehensive solution for many of C3.ai’s use cases would require potentially combining a mix of internal tools, SaaS solutions, open source, and legacy data products, leveraging a systems integrator.
The firm product offerings span two main platforms: their core product, the C3 AI Suite and C3 AI Applications.
The AI suite is described as a “core technology, comprehensive application development and runtime environment designed for rapid design, development, and deployment of Enterprise AI applications of any type”, while the applications layer is built on top of the Suite, with industry and application specific turnkey AI solutions.
C3 is cloud provider agnostic, and can operate in the cloud, on premises, or in a hybrid cloud.
The claim is that C3.ai provides high value outcomes and short time to value (as little as four weeks). C3 says it speeds up development by a factor of 26X and reduces the amount of code needed by 99% to deploy.
C3 says it has invested nearly $800M in building out its Suite product.
As often for AI companies, there is some level of questioning about what part of their activity is actually AI, vs more traditional software, and C3 is no exception. See for example this piece in ZDNet: Dissecting C3.ai’s secret sauce: less about AI, more about fixing Hadoop
C3 charges a fixed annual fee for platform access based on the number of seats, perpetual or term licenses / subscriptions for the applications, and a services fee for implementation.
- (Note that C3.ai’s fiscal year runs from May through April.)
- FY20 revenue ended at $156.7M, up 71% YoY from FY19 (which ended at $91.6M in revenues). Growth accelerated substantially from 48% in FY19. This lands C3 on the upper end of recent IPOs in terms of growth, but in the lower half for revenue.
- Growth acceleration correlates with a substantial increase in sales and marketing spend (+$57M YoY), causing sales and marketing as a percentage of revenue to jump from 41% to 61%. R&D investment remains heavy at 41% of revenue.
- Blended gross margin improved from 67% to 75% (a solid software business margin), with margin improvements coming from both subscription and professional services.
- C3 was largely a software driven business, with subscription revenue representing 86% of total FY20 revenue, up 1% from FY19 (85%)
- As tends to the case in software IPOs, C3 is not profitable. EBIDTA loss as a percentage of revenue increased from 36M (39% of revenue) to 71.5M (46% of revenue)
- Free cash flow was -$64.1M in FY20, up 54% from -$41.7m in FY19.
Sales & Marketing
- Heavy investment in sales and marketing in FY20, tripling YoY brand marketing spend from $9.5M to $34.5M before cutting back early in FY21 due to covid.
- As an aggregate, sales and marketing as a percent of revenue grew from 41% in FY19 to 61% in FY20 (as other departmental spend remained in line with revenue growth). This has allowed them to greatly accelerate revenue growth in FY20.
- C3’s sales team is organized into a geography and industry matrixed direct sales organization, with a supporting forward deployed engineer organization (similar to Palantir) and additional channel distribution partnerships.
- C3 has entered strategic partnerships across oil and gas (Baker Hughes), defense (Raytheon), and technology (AWS, Google, IBM, MSFT, etc.). Baker Hughes is both a customer ($39.5M a year in FY20) and a cobranded reseller (on the hook for $11.5M in FY20). Partners offer marketing and reseller support.
- C3’s sales cycle is long – 5 to 9 months from start to close, but trending down from 13 months in FY16 with an additional follow on 5 to 16 week paid trial.
- Their typical contract lands as a 3 year enterprise agreement, paid annually. Federal contracts are on average 11 months long.
- The average contract value jumped from $1.2M in 2016 to $12.1M in 2020. With an average 3 year contract, annual revenue would land ~$4M per contract.
- C3 employs a typical “Land and Expand” strategy for their enterprise customers.
- The average ACV of 15 largest customers was $12.8M, and they have purchased an average of $26.1M in additional services and product subscriptions.
- C3 has expressed a desire to target both Fortune 500 companies and downstream towards mid-market companies, but at their levels of revenue concentration and ACV, the downstream traction is unclear.
- They have started to diversify revenue across industries substantially, and gained traction in financial services, aerospace & defense, and oil and gas (led by the massive Hughes Baker contract).
|Industry as % of Rev||FY18||FY20|
|Oil & Gas||1%||29%|
|Aerospace & Defense||3%||18%|
Revenues are concentrated in North America:
As a general AI platform, the company targets a wide swath of use cases across a number of sectors, from financial services to manufacturing to oil and gas. This gives them flexibility to tackle new adjacent use cases, but also runs the risk of not positioning themselves as the market leader in any particular segment or function.
As such, their purported market sizing aggregates a number of segmented markets, projected in the S1 as totalling to $174B in 2020, growing to $271M in 2024 with a 12% CAGR.
Funding & Ownership
C3.ai has raised $356M according to CB Insights across 6 rounds of fundraising, with several large rounds led by TPG Capital’s Nehal Raj. Other investors include Baker Hughes, BlackRock, FS Investors, Breyer Capital, Sutter Hill, Interwest, Makena Capital.
During the offering, stock will be split into Class A stock with 1 vote each, and Class B stock holding 50 votes each, nearly wholly controlled by Thomas Siebel.
The filing indicates that Siebel holds 75.8% of the combined voting power, derived from ownership of 97.9% of the Class B shares, and 33.9% of Class A shares. TPG owns the largest investor stake at 22.6%, and strategic partner Baker Hughes holds 15.1%.