Investing In AI: Leading VCs Share 10 Green Lights And 10 Things That Send Them Running

At Alumni Ventures, we are one of the top investors in the sector and have a dedicated AI Fund. We have developed a list of the top 10 things we look for in AI companies, plus 10 red flags that send us running.

In this blog, we also talk about 5 specific start-ups we love.  And, if you want to learn more, view our on-demand webinar featuring top VCs investing in AI like B Capital and M12 – Microsoft’s venture Capital Fund.

10 GREEN LIGHTS THAT VCs LOOK FOR

1.  Fix-It Mindset

When evaluating startups, a key question is whether the venture addresses a significant, real-world problem. Successful startups must identify pressing issues affecting individuals, communities, or industries and then solve them with innovative solutions.

So when we search for investments in AI, we’re seeking startups leveraging this tool for problem-solving — whether that’s in data analysis, content creation, predictive modeling, image and speech recognition, optimization, personalization, etc.

For example, one major problem is that the demand for data and computing power to support AI and Large Language Models (LLMs) is expected to surge by 200x in the next two years. AV’s portfolio company Enfabrica is tackling that by developing networking chips, systems, and software tailored for AI computing tasks. By leveraging Enfabrica’s chips, companies can reduce AI cluster costs by 50% while also enhancing scalability (10-fold), bandwidth (2-fold), and data speeds (75%).

2. Are You Really Using AI?

As the number of AI startups explodes, VCs have also seen the rise of wanna-bes using AI buzzwords and sporting “dot-ai” domain names. While the term “AI” is increasingly common, its precise meaning remains elusive. It’s an umbrella term covering various mathematical approaches to simulate intelligent behaviors.

As an example: Companies claiming they’re AI-centric when basically indexing data are misusing the term. Indexing is a fundamental function performed by conventional software that has existed for decades. So evaluating AI startups means first assessing whether they genuinely incorporate AI or employ the term as a marketing ploy.

3. It’s the People, Stupid

You can’t overstate the value of an experienced, dynamic team in judging a startup. While a promising idea is critical, the team behind that idea can ultimately determine success. An experienced team brings knowledge, skills, and an ability to adapt and pivot when conditions change or challenges arise.

In evaluating an AI team, we want to know if they have a Chief Data Scientist and / or an engineering team with AI expertise and a track record of work in the field? It’s important to keep in mind that if a key talent is missing, the competition for top AI professionals is fierce.

We also drill down into the team’s combination of technical expertise, problem-solving skills, and essential personal qualities. Examples of what we’re looking for include

• Proficiency in programming languages commonly used in AI, such as Python
• Familiarity with relevant libraries and frameworks (e.g., TensorFlow and PyTorch)
• Solid foundation in mathematics, including linear algebra, calculus, probability, and statistics, which are fundamental to AI and machine learning
• Deep understanding of machine learning algorithms, including supervised and unsupervised learning, reinforcement learning, and deep learning.

A prime example of an AI team that checks the boxes: AV portfolio company ChainML boasts a leadership with extensive expertise, technical excellence, and a track record in AI and related domains. CEO Ron Bodkin brings 15 years of AI experience, with a leadership career in applied AI through Google’s CTO Office, as VP of AI Engineering & CIO at Vector Institute, and as Co-Founder of an AI company that Teradata acquired. He’s supported by a team of seasoned executives with deep experience in AI, scalable computing engineering, product development in applied AI and data science, and expertise in AI privacy and time series analysis.

4. Tech Chops

AI startups must demonstrate a deep understanding of AI technologies and their practical application. So what does a solid technical foundation for AI look like? We believe it demands a broad range of knowledge and skills across areas such as machine learning (ML), deep learning, optimization techniques, natural language processing, computer vision, etc.

And application is everything. For example, a “good machine learning model” is one that performs effectively and efficiently for a given task, producing accurate predictions or classifications while generalizing well to unseen data. Note that qualifier, “for a given task.” A model can vary greatly depending on the use case, domain, and objectives. A good model for medical diagnosis may require different criteria than one for movie recommendations, for instance.

5. Growing the Business

Is the product or service designed for cost-effective growth? Scalability isn’t just a feature but a strategy underpinning long-term value creation. Startups that understand the importance of designing their offerings to cost-effectively expand and adapt to increasing demand are more likely to succeed.

For instance, when it comes to a startup’s machine learning models, we want to understand the computing power demands of their operations and determine whether such power is readily accessible.

Escalating hardware requirements have prompted many startups to embrace cloud computing as a cost-effective solution. At the same time, cloud service providers are striving to deliver optimal hardware capabilities. Gaining insight into the resource requirements helps us evaluate a startup’s scalability potential.

6. Moats and More Moats

A competitive moat serves the same purpose as a castle moat: A durable advantage that helps protect the startup from invaders over time. Moats can take various forms, such as proprietary technology, a strong brand, network effects, exclusive partnerships, or regulatory barriers. Startups with a well-defined competitive moat are better positioned to withstand competition and become industry leaders.

Questions that come into play when evaluating the moats of an application-layer AI startup reliant on data:

• Is the data source unique?
• Will the company have access to the data over time?
• How does it plan to collect the data?
• How is the company quantifying data to train their AI models, and can these advanced machines self-improve over time?

If training datasets are public or easily accessible, the startup may have a limited competitive advantage. In contrast, a company that can generate proprietary training data has a more defensible position.

An AI startup with impressive moats is AV’s portfolio company Lasso, a natural language search engine for Web3 data. Lasso’s competitive advantage lies in its exclusive user-generated content database, with user-submitted questions, corresponding queries, and associated metadata. The dataset moat enables Lasso to enhance their models in ways that’s hard for competitors to replicate. Notably, this data holds high relevance to Lasso’s target audience, setting it apart from general search and discovery platforms — and creates a network effect where a growing user base leads to an expanded dataset and improved results.

7. Opportunity Size

To deliver meaningful returns on investment, a startup must target a substantial market. A large market offers the potential for exponential growth and, therefore, increased investor value. While startups that focus on niche markets might find success, their ultimate returns could be limited. Therefore, we prioritize startups that target growing, sizable markets, as these ventures are more likely to achieve meaningful market penetration and sustainable growth over time.

Drug discovery is one example of a massive, growing market sector powered by AI. In 2022, the worldwide drug discovery market was assessed at $55.46 billion. Projections indicate that it will reach ~$133.11 billion by 2032, with a compound annual growth rate (CAGR) of 9.2% from 2023 to 2032. One of AV’s portfolio companies, Iambic Therapeutics, is successfully targeting the sector with a biotech platform that harnesses the power of AI and quantum mechanics to transform drug discovery.

8. Making Headway

Traction or momentum — which might be in user engagement, strategic partnerships, or revenue generation — is a good indicator of a business’s potential. Traction signals that the startup is on the right path, creating a functional, viable business with the potential for scalability and long-term success.

AV’s portfolio company Cohere illustrates the principle. Considered the leading generative AI / natural language processing player for enterprise customers, the company has recently become a unicorn. Founded by several teammates from the cutting-edge Google Brain team, Cohere has achieved tremendous customer traction. That includes customers such as McKinsey, Salesforce, AWS, Jasper AI, Spotify, and more — plus partnership traction with leading cloud providers Google Cloud and AWS for unique access to compute power. Overall, we are impressed by its month-over-month user and usage growth.

9. Business Model = Money Engine

Beyond having a great product or service, a well-defined business model outlines how the company will actually bring it to market and make money on that offering. A clear, viable business model lets you understand how the startup

• Plans to generate revenue, achieve profitability, and ultimately scale
• Address customer acquisition, pricing strategies, cost structures, and revenue streams
• Has thought through its path to sustainability and growth.

10. Managing Money

A startup’s financial health and runway (a projection of how much time it has before it runs out of money, based on monthly expenditures) indicate its resilience and long-term viability. Startups with a clear financial runway have the funding to sustain their operations, avoid crippling cash-flow issues, and grow as planned.

Earlier, we referred to AI startup’s challenge of securing computing power — is it technically feasible to acquire what’s needed? For most AI companies, this is also a finance issue. Evaluating costs from a data-access standpoint also prompts us to examine whether the startup chooses cost-efficient, pre-trained machine learning models or custom trains, which provide greater flexibility but require more capital.

10 RED FLAGS THAT SEND VCs RUNNING

This section highlights 10 warning signs that apply to any startup, not just AI.  But we are seeing more of this normal given the spotlight on the space.

1. Hand Waving

Poorly defined business plans or technologies raise red flags. Startups must have a clear and well-defined business plan that outlines their strategy, target market, revenue model, and growth path. Similarly, the startup’s technology should be explained clearly and completely. Lack of clarity can confuse potential investors and hinder the startup’s communication of its value proposition in the marketplace. “Fuzziness” also suggests a potential risk of mismanagement or insufficient market understanding, which can lead to challenges in execution.

2. Crazy Pricing

A startup boasting a very high valuation without convincing evidence or performance metrics can trigger alarm bells for investors. Investors seek a balance between valuation and a startup’s actual performance and market potential. An inflated valuation can indicate unrealistic expectations or overhyped market perceptions, creating challenges in securing additional funding, achieving profitability, or sustaining long-term growth.

3. Rookies Acting Like Rookies

Startups led by inexperienced teams are more likely to make fundamental mistakes. Entrepreneurship is a challenging journey that demands a deep understanding of the industry, market dynamics, and business complexities. Experienced teams bring a wealth of knowledge, problem-solving skills, and the ability to navigate the inevitable hurdles of building a company. Inexperienced teams may need help with decision-making, strategic planning, and managing challenges, making them more vulnerable to costly errors.

4. Looking for a Problem to Fix

Evaluating startups from a product-market fit perspective is a basic test. Having a fantastic product is a significant achievement, but without market demand, it’s like a ship with no clear destination. A product must address real pain points and provide real value to customers. Achieving a strong product-market fit ensures that the “ship” has a destination and is on a prosperous course to reach it.

5. Fishy Behavior

Legal or regulatory red flags can be a deal-breaker. In evaluating startups, regulations can be major concerns, impacting the feasibility of an investment. Regulatory challenges can lead to delays, costly disputes, fines, or even shutdowns —  all of which can erode the value of an investment.

6. No Customers to Be Found

Low user engagement or poor revenue growth can be concerning indicators. A startup’s ability to attract, retain, and monetize customers are essential for its long-term success. Lack of traction may signify poor product-market fit, marketing strategy, or competitive challenges. In fact, it can often be fatal, impacting a startup’s ability to secure future funding, expand market share, or even sustain operations.

7. Spending Addiction

A startup that burns through its capital too quickly can be a high-risk proposition and raise concerns for investors about financial sustainability and resource management. Startups are expected to balance their growth ambitions with prudent financial management to demonstrate their ability to deliver long-term value. Investors typically seek startups with a clear and well-managed financial strategy, where capital is deployed efficiently to support growth and innovation.

8. Calling Bull%$^

Startups that prioritize generating hype rather than substantive product development can pose risks. While marketing and buzz are help build a brand and customer following, they should not come at the expense of creating a viable, sustainable business. Hype can lead to inflated expectations and a potential gap between what a startup promises and what it can deliver — ultimately disappointing customers, eroding trust, and hindering long-term prospects.

9. Going Dark

In any partnership with investors or customers, honesty and openness are nonnegotiable. Investors rely on accurate, timely, and candid information to make informed decisions, and customers expect clarity about the products or services they are engaging with. A startup that’s not sufficiently transparent may raise concerns about its integrity, competence, and long-term commitment to good community relationships.

10. Instant Gratification

Startups prioritizing immediate, short-term gains at the expense of long-term strategy often risk missing the bigger picture. Sustainable success in the business world requires a well-rounded approach that encompasses short-term profitability and long-term growth, market leadership, and adaptability to changing circumstances. Focusing on quick wins can result in hasty decisions and neglect investments in research, development, and infrastructure that are crucial for long-term value.

Your AI Investing Partner

AI presents one of the most exciting technology shifts we’ve seen in a generation. It’s groundbreaking in its applications, reach, and potential impact — analogous to the Internet and smartphone, in our estimation.

As one of the most active AI investors in the US, our team has seen and backed many AI companies. If you’d like a partner in navigating this complex landscape, we invite you to invest with us and “own a piece of the future.”