The Chief Executive of IBM has just announced that there is no chance that the current AI data center investments will give sufficient returns, shaking venture capital markets. This contrarian perspective of a traditional technology giant confirms our rigorous methodology of capital management in possibly overvalued areas as a venture capital firm.

The AI Capex Bubble in their parts.

The figures are mind-blowing: hyperscalers are planning spending in the range of 200B+ on AI infrastructure and the annual revenue of enterprise AI usage in the range of less than 20B. This is a 10:1 spending to revenue ratio that is reminiscent of historic bubbles. This has been the case with our venture capital firm in crypto mining (2018) and fiber optics (2000). This time it is speed–this bubble is fomenting in months not years.

The sustainability point is appealing to our discussion of venture capital in space economy startups. Space projects demand huge initial capital, yet are constrained in practice by their physical (orbital rights, launch windows, spectrum rights) capacity which inherently inhibits competition. The AI infrastructure is not subject to these limitations, which will generate a possible supply glut that will collapse returns.

Venture Capital Firm Strategy: Be Efficient.

Instead of shunning AI altogether, advanced venture capitalist companies are shifting to efficiency games. Our portfolio construction has now focused on:

1.     Capital-Light AI Applications: Startups that used the pre-trained models through API instead of developing their own infrastructure. By using existing models to analyze legal documents, one of the portfolio companies reached a 10M ARR with only 2M capital.
2.     Job assisting Technologies: Firms that have made AI more efficient -model compression, quantization, specialized inference chips. The IBM caution is actually an advantage of such start ups since businesses are looking at cost savings.
3.     Vertical integration: AI apps that are vertically incorporated within the high-margin sectors (biotech R&D, financial trading, semiconductor design) where the ROI calculations are calculated even at the present expenses.

Requested Adjustments in Due Diligence Framework.

Our venture capital company has made due diligence more rigorous in order to eliminate AI tourists among real builders:

Unit Economics Scrutiny: Portfolio companies should no longer be expected to meet positive unit economics at sub-$5M ARR, as they come to a later stage in life. The warning issued by IBM implies that numerous AI startups that grow their activities on the basis of negative unit economics will be in existential crisis when the capital markets contract.

Technical Moat Validation: “proprietary models” are no longer sufficient in their vagueness. Adversarial testing is done by our technical partners where startup outputs are compared against publicly available models. Where performance benefits do not equal higher costs, just move on.

Customer ROI Assurance: Customers Customer ROI: This is performed through direct interview with the customer to confirm the alleged ROI. A single startup in the category of AI sales enablement had very promising metrics until we found out that they were being used to create simple email templates rather than more advanced use cases they were being promoted.

A Counter-Cyclical Game by the Space Economy.

Interestingly, the investment venture capital of space economy startups give a balanced point of view. Space projects have long development cycles of several decades, which require capital discipline at the initial stage. It is in contrast to the culture of AI of going fast and spending money, which explains why space economy investments now constitute 15 percent of our new commitments, a substantial rise compared to 5 percent a couple years ago.

A Venture Capital firm should strike a balance between fast-paced industry such as AI and long-term investments such as space technology. Such a barbell method makes the portfolio resilient to any sentiment change which might precipitate when IBM is sending alerts.

Evolve Venture Capital has sector-specific investment theses which are updated quarterly. Our Sector Outlook Reports describe our positioning of the capital efficiency trends. Recently we held a closed-door briefing of LPs on “AI Bubble Dynamics and Portfolio Defense”–Highlights are accessible to potential co-investors on request. Interested in it, contact our Investor Relations team.

Financial Advisor Insights

We Evolve Venture Capital financial advisory desk: The IBM warning summarizes our doubts of AI infrastructure ROI. We are recommending model optimization and multi-cloud arbitrage as areas of immediate 20 percent cost-cutting programs by portfolio companies. In the case of LPs, we are proposing a strategic cut in AI infrastructure investments between 25 and 15 percent of tech exposure, and then moving the capital to application-layer businesses that have a history of strong ROI. The example of venture capital to invest in the startups of the space economy is also an appealing option, as the level of capital intensity is coupled with actual barriers to entry and government-sponsored demand. Most importantly, we are now asking the AI investments to prove that they have a way to profitability given a half-cut in the availability of venture funds- this rigor will eliminate casualties and the survivors when the market marks down.

Contact Information:

• Website: www.evolvevcap.com
• Email: contact@evolvevcap.com
• Phone: +65 8181 4097