Innodata: Mispriced AI Data Engineering And LLM Training Story, Primed For Growth

Thesis

Innodata (NASDAQ:INOD) is an undercovered artificial intelligence (AI) data engineering company that is trading significantly below the valuations of comparable private AI companies like ScaleAI. Now down ~40% from its post-earnings highs, the company's current valuation and forward-expected parabolic growth trajectory present an opportunity for investors to become a part of this company before it is discovered by the masses.

The company is a key player in the AI dataset training playing field, with four of the top five largest technology companies in the world having selected Innodata as their data engineering provider. Over the past quarters, it has become increasingly clear that Innodata will be one of the companies selling the "picks and shovels" to the AI revolution.

Thematics

The generative AI market opportunity over the next 10 years is staggering. According to Innodata's 2023 investor presentation, the total generative AI market is expected to grow from approximately $40 billion in 2022 to over $1 trillion in 2030, representing a compound annual growth rate (CAGR) of around 42%.

Furthermore, the subset of the generative AI market that Innodata is directly competing for, generative AI IT-focused services, is expected to grow from $83 million in 2022 to $85 billion in 2032. This represents a 100% CAGR over 10 years.

Though this subset of the AI market is in its early stages, it's clear that the AI arms race has begun. According to Innodata's 1H 2023 investor presentation, the big five (Meta, Amazon, Google, Microsoft, and Apple) are already deploying billions of capital expenditure (CAPEX) to forge a path forward and discover emergent properties in the technology.

The Company

Innodata's applied AI program is the result of decades of expertise in unstructured data and is being used to support generative AI builders and adopters. Innodata provides both data engineering services and platforms to assist. For AI builders, some of Innodata's key services include data science, synthetic data generation, reinforcement learning (RLHF), LLM integration, and AI application development. For adopters, Innodata provides platforms such as Agility Media Intelligence, Annotate (for supervised ML), Document Intelligence, and Synodex Medical Data Analytics. The culmination of services & platforms, decades of experience in unstructured data, and subject matter experts give Innodata a leg up in the space.

The steady stream of new and expanded contracts serves as clear validation that Innodata's approach to data science surpasses that of competitors.

Taking on Companies like ScaleAI

ScaleAI is a company that I would consider a direct competitor to Innodata. They also provide products such as data generation, labeling, and fine-tuning. As you can see from the graphic above, In addition to AI Infrastructure, data generation, and labeling, Innodata also competes in the data & AI consulting and Industry Application spaces. This gives the company quite a few transferable skills allowing them to be a more valuable option to potential customers.

In my view, it's likely that Innodata has directly competed against ScaleAI for contracts. Furthermore, as you will see from the commentary in the next section discussing Innodata winning contracts directly against 17 competitors, I also believe that it's likely that Innodata has directly won business over ScaleAI.

Sweeping the Playing Field

Let's take a look at this quote from Innodata's CEO, Jack Abuhoff, on the company's Q2 2023 earnings call:

On July 18, we announced having won another new customer, another of the top five global tech elites. Similarly, we will be helping this new customer develop and train large language models. The customer shared with us that it had been extremely impressed with our pilots and selected us from a field of 17 competitors. The agreement we signed with this second new customer provides for up to eight million in spend for the remainder of 2023 under an initial program, although how much of this we recognize is dependent upon a number of factors, including how quickly we ramp up. Ramp-up aside, based on discussions with this customer, we believe we may potentially get to an annualized run rate of $15 million with this customer by the end of the year solely on this initial program. The agreement we signed with this customer is also a framework agreement, which enables its business units and product centers to easily assign new programs and scope to us. We have already begun discussions with this customer regarding new programs.

The expected run-rate for this same customer was increased from $15 million to $25 million just over a month later. In my view, this is a great example of Innodata's ability to secure and expand business agreements with some of the world's largest technology companies.

I do not believe that these contracts are a "one-and-done" thing. Companies like Amazon, Google, Meta, Apple, and Microsoft are committed to spending billions in this AI arms race.

Just last week, Amazon committed up to a $4 billion investment in Anthropic, clearly demonstrating their willingness to spend billions to compete in the generative AI space

Forward Opportunities

Moving forward into 2024 and beyond, the company has numerous contracts expected to materialize and reach maturity.

First, the contract announced in the company's August 29, 2023, press release is expected to potentially reach an annual run rate of $25 million by the end of 2023. This is an expanded agreement related to the Big Five contract announced in the company's July 18, 2023, press release, which was initially expected to potentially reach an annualized run rate of $15 million by the end of 2023.

Next, the deal announced on Innodata's Q2 2023 earnings call with another Big Five technology company has the potential to be even larger. The agreement commits $3.5 million to start with AI and LLM development, with Innodata stating that its customer had shared a vision for the initial program which, if fully realized, could contribute approximately $12 million per quarter at maturity. It's safe to assume that deals of this magnitude can take quite a while to ramp up. Historically, we've seen deals take approximately 12 months to fully ramp.

In Innodata's June 27, 2023, press release, the company announced that it had been selected by an existing Big Five customer, "the Cloud hyperscaler", to perform AI data annotation and LLM fine-tuning as a white-labeled service for its cloud and platform customers. This deal could prove to be one of the largest revenue contributors to date if fully realized, as it gives Innodata the ability to scale and distribute LLM services to a pool of tens of thousands of end customers of the Cloud services company. As of Innodata's Q2 2023 earnings call, nine end-customers have signed on for pilots through the white-label service that the company believes is highly likely to turn into booked business in the near term. One of these pilots has an anticipated booking value of over $1 million. Furthermore, the company has stated that they now have a pipeline of pilot programs in the works, which include a legal information company, one of the largest life insurance companies, a leading investment bank, and a leading commercial bank. Based on the positive feedback since the initial announcement of the white-label program, Innodata now expects one to two new pilots each week.

I believe that this development in the story is larger than many anticipate, potentially giving Innodata the ability to scale its products and services to new businesses at a rapid pace.

Financials

In my view, comparing the company's 2023 and 2022 revenues does not fully capture the organic growth that the company has experienced when excluding the large social media customer (which I believe was Twitter), which paused its funding due to a "significant management change". This happens to coincide with Elon Musk's takeover and immediate decision to not pay vendors and landlords.

In 2023, Innodata's top-line revenues have been lacking contributions from a previous agreement with this large social media platform. When excluding the approximately $7 million in revenue contributions from who I believe was Twitter in 1H 2022, Innodata's 1H 2023 revenues have grown by 12% year-over-year organically. Furthermore, this baseline of organic revenue growth also does not yet include contributions from the two major Big Five contracts mentioned earlier.

Earnings Power at Scale

I believe that CEO Jack Abuhoff laid out some solid long-term ambitions for the company in its FY 2021 results press release:

Leveraging trends seen in 2021, we are budgeting a 30% increase in year-over-year revenues in 2022, an acceleration from the 20% growth we achieved in 2021. Longer term, our business plan calls for approximately $200 million in revenues and approximately 30% EBITDA margins by 2025

Though it appears that this ambition may have been slightly delayed, In my view, $200 million in annual revenues and around 25-30% EBITDA margins within the next few years is a good idea for Innodata's business performance as it scales. The prospects of the company have gotten significantly better since then as seen from the numerous large contracts signed in 2023.

As Innodata scales up, I believe its operating leverage becomes more clear. The company has stated that its approximate adjusted EBITDA breakeven point is around $18.5 million per quarter in revenues, which means at current revenue levels, we should see significant adjusted EBIDTA acceleration with expected revenue growth.

The following chart helps give some perspective on how Innodata ramps up to what I believe are revenues at scale.

The figure of $80 million for FY 2023 will be described in more detail in the "Valuation of Competitors" section below. Regarding the two new contracts previously discussed, at maturity, one is expected to reach an annualized run rate of approximately $25 million, and the other, if fully realized, could contribute another $12 million in revenues per quarter. Using a baseline of $80 million in revenues for both FY 2022 and FY 2023, both contracts have the potential to increase revenues to around $153 million at maturity. This would represent revenue growth of 91% over FY 2023. What's even more promising is that the $153 million does not include any other potential contracts and expansions. It also excludes any business signed from the white-label program, which I believe has the potential to be massive. In my view, all of these opportunities combined show a clear path to the company achieving yearly revenues of $200 million.

These are all points to consider for now. I will be publishing a more detailed forward EPS forecast in future coverage of Innodata as the company re-accelerates its revenues in the coming quarters.

Valuation of Competitors

Let's take Innodata's most direct competitor, ScaleAI, for example. ScaleAI is believed to have annual revenues somewhere in the $200 million to $300 million range at the time of this writing. ScaleAI is believed to be earning revenues in the territory of Innodata's longer-term ambitions while being valued at $7.3 billion. Taking the midpoint of that revenue range, ScaleAI is believed to be currently valued at 29.2x its current annual revenues.

Let's take a conservative approach for a forecast of Innodata's FY 2023 revenues. The company achieved revenues of $19.6 million in Q2 '23 and $18.8 million in Q1 '23, which represents about a 4% growth quarter over quarter. Though I expect revenues to significantly re-accelerate in 2024, we'll use a 4% quarterly growth rate for the remainder of 2023. That puts us at around $80 million for FY 2023 and a price-to-sales ratio of 2.78. That's a forward price-to-sales ratio of 1.19 on $200 million in annual revenues per the company's previous expectations.

Sales growth from $80 million in FY 2023 to $200 million at scale represents an increase of 150%. This is on top of an underlying industry that is expected to grow at a CAGR of 100% through 2032.

Cashflow and Liquidity

The company has a clean balance sheet with no external debt or borrowings and had $13.6 million in cash and short-term investments at the end of Q2 2023. With the company's cash on hand, historical burn rate, and access to a Wells Fargo credit line of an additional $10 million, Innodata's liquidity position appears strong for the time being. Furthermore, the company has not raised any capital from external investors since its inception, funding all of its operations through organic cash flow.

Risks

Competition: Innodata has demonstrated a strong competitive advantage thus far, having been selected and validated by the Big Five tech companies from a large pool of competitors. Innodata must continue to leverage and expand upon its first-mover advantage while also continuing to demonstrate that its data engineering and applied AI program is superior.

Termination or loss of business from the Big Five: As the most significant drivers of revenue growth over the next year or so, any loss of these contracts would pose a threat to the bullish thesis on Innodata.

Insider selling: Recent insider selling has unfortunately, whether you agree with them or not, put those on the bearish side of this story in control of the narrative (and the stock price). According to a Schedule 13D from April 6, 2023, CEO Jack Abuhoff beneficially owns an aggregate of 2,995,024 shares. Jack sold 150,000 shares in August of 2023, which was about 5% of his beneficially owned shares. In my view, 5% of beneficially owned shares is not "dumping," as some have labeled the transaction.

The chairman of the board, Nauman (NICK) Toor, has sold approximately half of his holdings throughout the year. This level of selling from the chairman of the board is something that is hard to justify or overlook and sends the market a very negative signal.

Despite this level of selling from the chairman of the board, I do not think this undercuts the future prospects of the business.

What I'm Watching Moving Forward

Contract sustainability: I believe a key focus for investors must be contract sustainability. These large deals are transformative, but how these agreements will expand beyond the initial work must be monitored closely.

White-Label Business: Apart from recurring revenues from ongoing work, I expect the conversion of white-label pilots into long-term business to be a driver of sustained revenue growth. It's evident from the pool of potential customers that there's a lot of opportunity and ample work for Innodata in the coming years. These pilots must continue to convert into confirmed deals.

Revenue ramp: Based on the company's current agreements and the white-label program, I anticipate a significant increase in revenues around Q2 of 2024. This revenue ramp-up needs to materialize within the next 12-24 months.

Insider buying and selling: I would like to see Innodata put a 10B5 program in place in the near future. I believe this would be a great way for management to be more shareholder-friendly.

Conclusion

Based on current and anticipated business agreements, as well as the tailwinds from an industry segment expected to grow at 100% CAGR through 2032, Innodata has the potential to grow into something significantly larger than its current valuation suggests. The company appears to be a key player in the AI dataset annotation subset, all while being valued at a mere fraction of its competition. In my opinion, it is entirely feasible for Innodata to achieve quarterly revenues of $50+ million and exhibit substantial earnings potential at scale. Now, it falls on Innodata's management to deliver on the contracts already announced and continue building upon its competitive advantage and momentum.

I believe that Innodata shares present an enticing opportunity, especially considering that the stock has declined by approximately 40% from its post-earnings highs. For this reason, I will be consistently monitoring and re-evaluating the company's financials, valuation, contract sustainability, growth prospects, and risk factors on an ongoing basis. You can expect updates to my coverage of Innodata to be published on Seeking Alpha with each Innodata earnings release in the future. Thank you for reading.