From: allin
Current AI Market Frenzy
The AI frenzy in Silicon Valley continues, impacting funding from early to growth stages, as well as M&A events [00:51:06]. The space of AI infrastructure for enterprises is currently considered the hottest area in venture capital [00:53:04].
Strategic Acquisitions in AI Infrastructure
Databricks, a privately held data infrastructure company, announced the acquisition of MosaicML for 38 billion [00:51:27]. MosaicML, founded by the individual who started Nirvana (an early AI business acquired by Intel), offers open-source AI models [00:51:50]. Rumors suggest MosaicML’s Annual Recurring Revenue (ARR) grew from 20 million since January, though other estimates place it at 400 million post-money valuation, implying about $10 million in ARR [00:53:35]. This valuation was considered reasonable for a fast-growing company in a hot sector [00:53:50].
Another notable acquisition is Snowflake, a competitor to Databricks, acquiring Neva for $150 million [00:56:01]. These acquisitions indicate that for companies in the data infrastructure business, it is becoming critical to integrate AI capabilities into their core offerings [00:56:16]. The interpretation of data through models and the tooling to build these models are now essential components of the software toolkit that companies must provide to their clients [00:56:28]. This shift suggests that more acquisitions are likely to occur as companies seek to expand their capabilities [00:56:48].
The strategic importance of these acquisitions is rooted in building an end-to-end toolchain for enterprises [00:55:27]. Enterprises are looking to deploy their own versions of ChatGPT internally, allowing employees to query company data with appropriate permissions and security [01:07:08]. The core stack for rolling out such an internal model includes data capture, labeling, storage, selection of an open-source model (e.g., from Hugging Face), and then training or customizing that model [01:08:12]. This last mile problem of customizing open-source models for specific enterprise needs is a major area of activity [01:08:50]. A significant challenge in this ecosystem is the current shortage of GPUs, which is not expected to improve for one to two years [00:55:40].
Large AI Funding Rounds and Capital Expenditure
Inflection AI, co-founded by Mustafa Suleyman (DeepMind co-founder) and Reid Hoffman, recently closed a 40,000 to 900 million [01:01:50]. This means approximately 1.5 billion raised is allocated to capex, leaving about 500 million for SG&A (Selling, General, and Administrative expenses) and about 2.5 billion of enterprise value attributed to their chatbot, Pi [01:02:17].
Concerns were raised about the high proportion of funding allocated to capex for a startup [01:03:39]. When a chipmaker (Nvidia) and a cloud provider (Microsoft) contribute significant funding, it can be viewed as “round-tripping cash,” where the investors provide money that is then used to purchase their services, effectively boosting their own revenue [01:04:02].
Contrasting Trends: Past Hype Cycles and Due Diligence
The current AI frenzy and its associated high valuations are a recurring pattern in Silicon Valley, previously seen with crypto, co-working spaces, synthetic biology, and SaaS [01:10:53]. This phenomenon is often attributed to a lack of checks and balances and inexperienced individuals in key roles, particularly within venture capital firms [01:11:21].
The shutdown of social messaging startup IRL, which was valued at over 170 million (led by SoftBank’s Vision Fund), highlights these issues [01:09:51]. An investigation revealed that 95% of its claimed 20 million users were fake [01:09:47]. Similarly, Byju’s, an Indian ed-tech startup once valued at 5.1 billion [01:10:17].
These cases underscore the importance of due diligence in venture capital [01:13:39]. Inexperienced board members and venture capitalists may fail to ask basic, uncomfortable questions or verify data, leading to inexcusable outcomes [01:12:08]. During periods of market froth, founders might leverage high demand to avoid thorough due diligence, with some investors suspending disbelief and assigning high-growth, high-gross-margin valuations to businesses that are fundamentally low-margin [01:15:52].
The size of venture funds also plays a role in these dynamics [01:17:35]. Large funds, such as SoftBank’s 200 million) to justify their management time, even if the target companies are effectively seed or Series A stage [01:18:08]. This leads to mistakes that are disproportionately large compared to optimal fund sizes (suggested to be between 600 million) [01:18:17]. The increasing size of funds is driven by the desire for higher annual management fees, potentially at the expense of sound investment decisions and thorough diligence [01:20:49].
As a result, fundraising for late-stage funds has seen a significant downturn, with firms like Insight and Tiger Global experiencing 80-90% reductions in their target fund sizes [01:22:17]. This suggests a “sovereign crunch” in late-stage financing, which is expected to worsen [01:22:57]. It’s estimated that a large percentage of “unicorn” companies (perhaps 60-70%) may turn out to be “zombie-corns” that go to zero, with many others facing down rounds, leading to a blended return for this era of roughly 1.1x [01:23:09]. The market is sending a clear message that this time is not different from previous hype cycles [01:24:08].