# AI Buildout Enters Riskier Phase, Says Parnassus CIO _Parnassus Investments CIO Todd Ahlsten discusses the AI buildout's risks, highlighting bottlenecks, GPU economic life, and software sector vulnerabilities._ **Published:** 2026-08-17 **Source:** https://www.startuphub.ai/ai-news/investors-news/2026/ai-buildout-enters-riskier-phase-says-parnassus-cio --- The current wave of investment into AI infrastructure is creating significant bottlenecks, with the physical world struggling to keep pace with digital demand, according to Todd Ahlsten, Chief Investment Officer and Portfolio Manager at Parnassus Investments. Speaking on Bloomberg TV, Ahlsten cautioned that while recent announcements from major players like [Nvidia (NASDAQ:NVDA)](https://www.google.com/finance/quote/NVDA:NASDAQ), Anthropic, and SpaceX are generating excitement, investors need to be discerning and look beyond the immediate hype. AI Buildout RisksDriver Parnassus CIO Todd Ahlsten highlights increasing risks in the AI infrastructure buildout phaseFrom the article 2 mentionsHe drew a parallel to the national deficit and the need to fund housing, suggesting that increased borrowing for infrastructure could lead to crowding out and increased risk.Unprecedented AI CapexCoreFrom the articleAhlsten highlighted that the unprecedented capital expenditure in AI compute is leading to bottlenecks across the industry.GPU Economic LifeDriverconcerns about the actual economic life and rapid obsolescence of GPUsFrom the articleAhlsten questioned the modeling of a GPU's economic life in such financing deals, pointing out the difference between depreciation and actual utility.Software Sector VulnerableDrivervulnerabilities in the software sector due to rapid changes and intense competitionFrom the articleConversely, Ahlsten identified software companies like Salesforce, Workday, and ServiceNow as potentially vulnerable.causesBottlenecks EmergeDriverFrom the article 2 mentionsThe current wave of investment into AI infrastructure is creating significant bottlenecks, with the physical world struggling to keep pace with digital demand, according to Todd Ahlsten, Chief Investment Officer and Portfolio Manager at Parnassus Investments.requiresDiscerning InvestmentContextinvestors need a balanced approach, looking beyond immediate hype for long-term valueFrom the article 4 mentionsSpeaking on Bloomberg TV, Ahlsten cautioned that while recent announcements from major players like Nvidia (NASDAQ:NVDA), Anthropic, and SpaceX are generating excitement, investors need to be discerning and look beyond the immediate hype.leads toIdentify Long-Term WinnersEffectfocus on companies with sustainable advantages and real revenue, not just excitementFrom the articleHe pointed to companies like Vulcan Materials, which supply concrete for data center construction, and Linde, which provides industrial gases with long-term take-or-pay agreements, as examples of second and third-order winners. ## The AI Infrastructure Boom and Its Bottlenecks Ahlsten highlighted that the unprecedented capital expenditure in AI compute is leading to bottlenecks across the industry. He noted that companies like Anthropic are reporting real revenue and updated ARR figures, which are moving the market for chipmakers and software providers alike. However, he stressed the importance of a balanced investment approach, urging investors to avoid getting "swept up" and instead focus on identifying long-term beneficiaries. The full discussion can be found on **Bloomberg Technology**'s YouTube channel. ![](https://img.youtube.com/vi/dwm_Q5ckOdU/maxresdefault.jpg) AI Buildout Entering a Riskier Phase, Says Parnassus CIO, from Bloomberg Technology A significant point of discussion was [Nvidia's](https://www.google.com/finance/quote/NVDA:NASDAQ) recent announcement of a $500 billion initiative involving six Wall Street firms to channel third-party capital. Ahlsten questioned the modeling of a GPU's economic life in such financing deals, pointing out the difference between depreciation and actual utility. He believes that GPUs are evolving into an asset class of their own, but raised concerns about the durability of these assets given the rapid pace of innovation and the potential for new architectures and memory solutions that could render current hardware obsolete. ## Assessing Risk and Return in AI Investments Ahlsten suggested that while the next two to three years for AI infrastructure investments might be predictable, the five to ten-year horizon presents significant uncertainty. He advised investors to carefully consider the risk-adjusted returns needed to participate in this dynamic market. The conversation also touched upon the broader market impact, with Alphabet Inc. (NASDAQ:GOOGL) seeking capital through equity and bond markets. Ahlsten viewed this as a sign that the market is entering a riskier phase of the cycle. He drew a parallel to the national deficit and the need to fund housing, suggesting that increased borrowing for infrastructure could lead to crowding out and increased risk. Regarding the concern of circular financing, particularly with companies like [Nvidia](https://www.google.com/finance/quote/NVDA:NASDAQ) investing its own capital into projects where customers lease compute, Ahlsten emphasized the need for investors to identify opportunities that mitigate leverage and risk. He pointed to companies like Vulcan Materials, which supply concrete for data center construction, and Linde, which provides industrial gases with long-term take-or-pay agreements, as examples of second and third-order winners. ## Identifying Winners and Those at Risk Ahlsten expressed confidence in companies like [Nvidia](https://www.google.com/finance/quote/NVDA:NASDAQ) and [AMD (NASDAQ:AMD)](https://www.google.com/finance/quote/AMD:NASDAQ), praising their technological advancements and market positioning. He noted that AMD's Helios 450 ramp appears promising, and its diverse compute offerings (FPGAs, CPUs, GPUs) make it a strong contender. He also highlighted that AMD is less involved in the direct customer financing that could raise circular financing concerns. Conversely, Ahlsten identified software companies like Salesforce, Workday, and ServiceNow as potentially vulnerable. He argued that their traditional seat-license models are under increasing pressure. The rise of token-based AI usage, as seen with [Google's (NASDAQ:GOOGL)](https://www.google.com/finance/quote/GOOGL:NASDAQ) Gemini and other AI models, could disrupt these established software giants. Ahlsten believes the pricing and usage dynamics of AI services will challenge the long-term bankability of the seat-license model, even for well-established companies. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.