Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
The Wave of Innovation: The "SaaS Apocalypse" narrative is an expected reaction to a new wave of innovation [00:00:05]. Historically, shifts like the transition to the hosted cloud faced similar skepticism that it would eliminate enterprise software; instead, the cloud ushered in its greatest value expansion [00:00:40].
Enterprise Software Longevity: Enterprise software remains the most valuable business tool introduced in our economy in the last 50 years and likely for the next 50 years [00:00:48].
The Transition to Agentic Systems: Enterprise software companies will split into two states: they will either become "agentic" or adapt to a new standard called the Rule of 70 [00:01:03]. The Rule of 70 is an AI-driven expansion of the traditional SaaS "Rule of 40," combining accelerated growth rates and expanding EBITDA margins [00:01:10].
Companies with "No Right to Exist": Roughly 8% of Vista's portfolio fell into a category of companies that have "no right to exist" under the new AI paradigm [00:02:46]. These are businesses that merely aggregate data from public sources and build analytics or "pretty pictures" around it [00:01:38]. Because Large Language Models (LLMs) inherently absorb public data, these standalone businesses face obsolescence [00:01:47]. The remaining 92% have a viable path to evolving [00:02:50].
Data Architecture: Models to Data vs. Data to Models
Data Leeching: Bringing enterprise data out of a secure corporate environment and sending it to external models "leeches" enterprise value [00:02:11].
The 2026 Mandate: Citing Satya Nadella’s remarks at Davos in 2026, the primary focus for executives must be bringing AI models to the data within their own secure environments, rather than sending proprietary data to external models [00:02:06]. This allows enterprises to preserve their intellectual property, context, trust, and security—often utilizing air-gapped environments [00:02:18, 00:09:38].
Enterprise Data Training Deficit: Less than 1% of actual enterprise data has been used to train existing LLMs [00:01:52].
The Vault Analogy: Allowing external entities access to raw corporate data is akin to opening a vault of proprietary insights to a newly minted armored car, risking leakage to competitors via optimized prompting [00:09:07].
The Evolution of Corporate Migration "Factories"
The Cloud Factory (2013): Over 12 years ago (around 2013), Vista recognized the shift from on-premise to cloud software [00:03:16]. Due to a shortage of distributed compute, Robert F. Smith struck a strategic deal with Andy Jassy (when AWS was early in its scaling phase) to guarantee compute capacity [00:03:24]. Eighteen months later, Satya Nadella introduced Azure, prompting Vista to build a second factory line [00:04:20]. This structural shift delivered a 2.5x economic rent pickup for converted portfolio companies [00:04:29].
The Agentic Factory: Vista built an "Agentic Factory" and "Agentic Lab" over the last 3.5 to 4 years to transition companies from cloud to agentic systems [00:03:02, 00:04:36]. Vista has put 54 of its 93 portfolio companies through this factory [00:06:12].
Value Inflexion: Instead of a 2.5x pickup, agentic systems deliver an order of magnitude higher terminal value [00:04:53]. For example, a company generating $400,000 in traditional SaaS revenue can layer $500,000 in agentic infrastructure components, ultimately terminating at a total valuation/value of $2.5 million to $3 million due to increased economic rent captured for the end customer [00:05:04].
The Economics of Inference and Infrastructure (DC2)
The Four Agentic Personas: Vista's research identified four primary archetypes of corporate AI agents, each exhibiting a distinct inference consumption pattern [00:06:09]:
The Worker That Never Sleeps: Continuously measures metrics every single day [00:06:24].
The Smart Sidekick: An everyday assistive tool used regularly [00:06:29].
The Heavy Lift: Built for intensive, heavy workload execution [00:06:32].
The Orchestrator: Coordinates, manages, and spins up multiple sub-agents [00:06:34].
The Cost Extraction Problem: The narrow group of primary inference providers can hike token prices up to 6x, extracting the margin and economic rent created by agentic software [00:06:44].
SambaNova Partnership: To combat token cost extraction, Vista partnered with SambaNova (led by CEO Rodrigo) [00:07:18]. Smith showcases an air-cooled (non-liquid cooled) chip that operates at 1/5th the power consumption while delivering 7x to 20x the throughput of legacy platforms [00:07:24].
Inference Cloud (DC2): Using this hardware, Vista converts existing brownfield data centers into "DC2"—a proprietary inference cloud [00:07:35, 00:07:55]. This ensures low-cost token capabilities that preserve gross margins for their over 90 portfolio companies [00:06:01, 00:07:43].
Misconceptions of AI in Enterprise: Probabilistic vs. Deterministic
Precision Over Probability: The biggest misconception is that AI can seamlessly handle all operations out of the box [00:08:05]. AI models are intrinsically probabilistic systems, whereas enterprise workflows demand deterministic, binary, and precise outcomes [00:08:13]. In a corporate environment, core operations like a wire transfer cannot be "mostly right" [00:08:35].
The Scale of the Model: Enterprises do not always need massive, expensive LLMs built by armies of PhDs [00:15:59]. For localized tasks—such as processing insurance claims or municipal pothole tracking—smaller, highly context-specific models ("five really smart high schoolers") are faster, safer, and cheaper [00:16:06].
Highly Contextual Sovereignty: Software serving highly regulated or hyper-local operational workflows possesses "dominion and sovereignty" over unique datasets [00:13:15, 00:14:18]. Example: Vista’s portfolio company ESO, which manages emergency response checklists for ambulances, tracks physical local data to verify oxygen levels before deployment [00:13:48]. These specialized businesses are highly defensible and structurally poised to expand [00:14:20].
Fund Deployment, Capital Markets, and IPOs
Deployment Slowdown: Vista, which raised a $20 billion fund, intentionally slowed its capital deployment pace over the last 12 to 24 months [00:09:54]. This pause allowed them to build out the underlying inference infrastructure (DC2) and underwriting frameworks necessary to ensure that portfolio companies retain the economic rent of AI features rather than sacrificing them to compute costs [00:10:08, 00:12:32].
Underwriting Shifts: Underwriting has shifted from static "stock" metrics to managing "flow"—evaluating a software company's capacity to build, secure, and run cost-efficient agents [00:11:00].
The IPO Landscape: Commenting on the confidential filings of OpenAI, Anthropic, and the public entry of SpaceX, Smith notes the market dynamic is heavily dictated by "SpaceX and then everything else" [00:14:33, 00:15:04]. Citing Capital Markets lead Ashley McNeill (formerly Morgan Stanley), the sheer size and volume of these listings reshape the public landscape [00:14:56]. Smith observed a split sentiment where current private holders are selling while external buyers are scrambling to get in, largely because "nobody wants to bet against Elon" [00:15:21].
Historical Reflections and the Future of Work
A New Industrial Revolution: Having started his career as an intern at Bell Labs working on semiconductor testing, Smith considers the current AI revolution to be the most profound technological shift he has witnessed—surpassing even the impact of the transistor or the printing press [00:17:08, 00:17:23].
Anecdote on Technology Scaling: Reflecting on technical complexity, Smith noted that his early career involved testing seven-layer circuit boards, compared to the 42-layer boards utilized for the modern SambaNova hardware instance he displays [00:18:15].
Impact on Labor: Agents are workers designed to perform tasks [00:18:33]. While tasks aggregate into jobs—implying a massive structural impact and reshaping of the labor equilibrium—this transition represents an evolution rather than total displacement [00:18:37].
Preserving Internships: Smith strongly cautions companies against dismantling human internship programs in favor of total automation, emphasizing that internships are critical for fostering optimism, creating a new group of thinkers, and driving long-term innovation [00:18:52].
The Future of Software in One Word: "Bright" [00:19:16].
Jul 16, 2026
Secrets of building The Whole Truth | Shashank Mehta, Founder and CEO | Unstarted | 16 Jul 2026 | Z47 Moments
"I fundamentally cannot live with the gap between my do and my say i find hypocrisy very very putting off" Shashank Mehta 07:04 https://youtu.be/HA7kNZgkcT8?si=CyHcafj8CzT5cQBu&t=7m4s "if you craft your life around your weaknesses you will…