"I think generative AI is at its heart con... and seeing these ultra rich ultra powerful people lie through their teeth turns my stomach." - Ed Zitron [00:00:00]
"This is the largest non-consensual push of technology in history." - Ed Zitron [00:00:26]
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
"OpenAI lost $20.9 billion last year... None of these people can just say 'Yeah we're on the path to making this profitable' because they can't." - Ed Zitron [00:00:44]
"The wrong hands are the hands of those who are running these companies. We should not be training these models to do these things." - Ed Zitron [01:15:28]
"It makes the easy things easy, the hard things harder." - Karl Brown (quoted by Ed Zitron) [01:30:05]
"These companies have gone from being cash machines to cash furnaces... adding more than $700 billion of new property, plants, and equipment in the last four years." - Ed Zitron [01:52:13]
"Buying AI GPUs allows them to kick the can further... allows them to say 'We're still doing something, we're working on AI, don't think too hard.'" - Ed Zitron [02:13:00]
Speakers & Credentials
Steven Bartlett (Host): Entrepreneur, investor, author, and host of The Diary Of A CEO, one of Europe's most listened-to business podcasts.
Ed Zitron (Guest): Tech critic, PR agency founder (EZPR), newsletter writer (Where's Your Ed At), author, and host of the podcast Better Offline. Known for his rigorous investigative writing on tech economics, corporate financial disclosures, and critical analysis of Silicon Valley hype cycles.
1. Executive Summary
Generative AI is fundamentally an unsustainable financial and technological "con" driven by disingenuous corporate marketing rather than genuine software breakthroughs 00:02:57.
The primary generative AI powerhouses—OpenAI and Anthropic—are massively unprofitable entities that only exist due to circular capital infusions from Big Tech giants like Microsoft, Amazon, and Google 00:04:48.
Consumers and enterprises do not pay the true cost of token generation; user pricing is heavily subsidized through loss-leading fixed subscriptions, masking extreme underlying computing costs 00:13:02.
Scaling large language models (LLMs) faces severe diminishing returns, power infrastructure limits, and hardware constraints, contradicting claims of imminent Artificial General Intelligence (AGI) 01:36:36.
Generative AI is being forced onto users non-consensually across major enterprise platforms to create artificial engagement and inflate queries 00:08:41.
Big Tech hyperscalers use massive AI capital expenditure (capex) spending to obscure slowing revenue growth in their core legacy business segments 01:46:17.
AI-assisted code generation is degrading global software quality, leading to measurable increases in technical outages and infrastructure crashes 00:23:50.
The market valuation of frontier AI companies relies on paper gains and circular debt financing that risk triggering a broad tech depression when capital flows dry up 02:11:05.
A financial collapse centered around OpenAI's inability to go public or sustain cash burn could occur around 2027, severely impacting broader equity markets 02:11:05.
01:28:13 - Commoditization of Content vs. Irreplaceable Human Taste
01:32:00 - AI Safety PR Hype vs. Modern Real-World Harms
01:36:36 - Technological Plateaus & The Hardware Breakthrough Requirement
01:45:12 - Financializing Tech: Zero Interest Rates to the "Rotcom" Bubble
01:49:34 - Capital Intensity & Financial Structure of AWS vs. OpenAI
01:55:23 - Big Tech Executive Rhetoric vs. Asset-Light Cash Furnaces
02:00:24 - The Myth of Autonomous AI Blackmail & PR Manipulations
02:08:31 - The 2027 AI Crash Timeline: OpenAI Cash Burn & Market Contagion
02:18:07 - Paper Gains, Private Credit, & Investor Actionable Warnings
02:25:00 - Closing Thoughts: Human Connection & Reclaiming Ground Truth
3. Detailed Thematic Summary
The "Con" Framework & The Circular Capital Illusion
Generative AI is not sold as standard enterprise software; it is marketed as a magical cure-all, masking an underlying model that is highly unprofitable and economically fragile 00:02:57. The financial foundation of the current AI boom is built on circular financing structures involving hyperscalers (Microsoft, Amazon, Google), hardware providers (Nvidia), and frontier labs (OpenAI, Anthropic) 00:04:35.
Approximately 70% of all global generative AI revenue is concentrated within OpenAI and Anthropic—two unviable entities that rely entirely on cash infusions from Big Tech 00:04:48. For instance, Amazon committed $50 billion to OpenAI and $5 billion to Anthropic, while Google directed $10 billion to Anthropic 00:04:56. Sellside analysts project that Big Tech expects over $400 billion in cumulative AI revenue to validate stock prices over the next three and a half years 00:05:11. However, outside OpenAI and Anthropic, total global software revenue directly tied to generative AI barely reaches $22 billion 00:19:15.
Public hyperscalers intentionally obscure their true AI revenues through unstandardized metrics like "annualized run rates," which fluctuate arbitrarily between 12-month, 13-month, or trailing 4-week multipliers 00:05:38. For fiscal year 2026, Microsoft reported $34.33 billion in cloud growth, but $24.1 billion of that figure came directly from OpenAI's compute usage 00:18:19. This leaves Microsoft with approximately $10 billion in organic revenue against $115 billion in capital expenditures in a single year 00:18:29.
The Unviable Unit Economics of Token Generation
The cost structure of large language models relies on token usage, where a token equals approximately 0.75 words 00:12:08. Charges occur for input tokens (prompts, raw data ingestion) and output tokens (responses and intermediate internal reasoning steps) 00:12:26. Consumer SaaS models ($20 to $200 per month) heavily subsidize heavy users, masking the actual costs 00:13:02.
SemiAnalysis research indicates that a user on a $200/month ChatGPT tier can consume up to $14,000 worth of compute tokens 00:13:02. Similarly, an Anthropic $200 tier user can burn through $8,000 in tokens, while a base $20/month tier can easily generate $400 in actual token costs 00:13:12. This pricing strategy led OpenAI to post a $20.9 billion net loss in a single year 00:13:26.
When frontier labs attempted to shift enterprise clients (>150 employees) toward pay-per-token pricing models, organizations experienced severe budget overruns 00:13:36. For example, Uber exhausted its entire annual AI token budget in just three months, forcing executive leadership to halt usage 00:13:55. Users pay for tokens regardless of whether output is accurate or incorrect, making code refactoring errors and hallucinations a billable event for the user 00:15:50.
Physical Infrastructure, Power Limits, & Stargate Abilene
Massive capital expenditures ($1+ trillion spent globally on capex, with another trillion planned) fund power-intensive hardware clusters that offer poor long-term utility 00:06:22. Unlike general-purpose CPUs or fiber optic lines, specialized AI GPUs offer minimal value outside machine learning workloads 00:09:42.
A prime example of this infrastructure strain is OpenAI and Oracle's "Stargate Abilene" facility in Abilene, Texas 00:07:04. The project draws 1.2 gigawatts (GW) of power, hosting 400,000 Nvidia GB200 GPUs distributed across eight specialized structures (50,000 GPUs per building) 00:07:12.
STARGATE ABILENE POWER & SPACE DENSITY COMPARISON
# Power Footprint (1.2 Gigawatts)
Stargate Abilene : [1.2 GW Condensed]
City of Bristol : [0.78 GW Citywide Total] [[07:22](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=442)]
# Physical Land Area (Square Feet)
Stargate Abilene : 998,000 sq ft [[07:31](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=451)]
City of Bristol : 1,200,000,000 sq ft [[07:31](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=451)]
(Stargate compresses city-level power into 1,172x less land)
This rapid buildout creates severe negative local externalities. Data center developers deploy unfiltered fossil-fuel gas turbines to bypass grid connection delays, creating localized air pollution 01:41:40. Additionally, heavy GPU production drives up memory costs and component prices across consumer electronics 00:42:25.
The push for widespread AI adoption has negatively impacted software development quality and digital ecosystem health 00:23:50. The surge in AI-generated code has correlated with a rise in technical outages, system downtime, and platform bugs across major tech companies 00:56:39.
Google, Microsoft, Amazon Web Services, and GitHub have all experienced increased service instability as automated code pushes overload existing infrastructure 00:56:04. GitHub has become inundated with unvetted code contributions generated by under-trained developers relying on LLMs 00:57:30. This shift encourages human complacency: developers review automated code less rigorously, compounding logic errors and security flaws over time 00:57:37.
AI CODE COMPLACENCY CYCLE
+-----------------------+
| LLM Generates Code |
+-----------------------+
|
v
+-----------------------+ Human assumes output is correct
| Human Skips Detailed | ------------------------------------+
| Code Review | |
+-----------------------+ |
| |
v v
+-----------------------+ +----------------+
| Flawed Code Pushed to | | Multiplicative |
| Open Source / Prod | | Technical Debt |
+-----------------------+ +----------------+
| ^
v |
+-----------------------+ |
| Infrastructure System | ------------------------------------+
| Outages & Bug Spikes | (e.g., AWS / GitHub Downtime Incidents)
+-----------------------+
Claims of falling error rates are often based on artificial benchmarks designed around basic summarization tasks 00:26:54. While benchmarks like the Vectara Hallucination Leaderboard show summarization errors dropping from 21.8% to 0.7%, real-world performance on complex tasks remains inconsistent 00:27:00. In production environments, LLMs struggle with contextual understanding and logic, leading to critical errors in financial models, legal research, and enterprise databases 00:25:33.
The degradation of modern web search stems from internal structural shifts within major tech platforms, later compounded by generative AI integration 00:53:08. In 2019, Google search engineers declared an internal "Code Yellow" due to flatlining user query growth 00:53:14. Ben Gomes (then Head of Search) and senior engineers argued that forcing artificial query growth would require serving worse answers, forcing users to search multiple times to find accurate information 00:53:30.
Despite internal pushback, Prabhakar Raghavan (then Head of Ads) advanced a growth strategy focused on maximizing ad impressions over result relevance 00:54:09. In early 2020, Raghavan replaced Gomes as Head of Search 00:54:23. Under his leadership, Google reduced anti-spam protections, allowing low-quality SEO content to rank higher and drive up total search volume 00:54:35.
When OpenAI launched ChatGPT, Google responded by deploying "AI Overviews" at the top of search results 00:54:58. This structural pivot aimed to keep users contained within Google's ecosystem rather than directing them to external websites, trading accuracy for user retention 00:55:14.
The 2027 AI Crash & Macroeconomic Contagion
A major market correction in the tech sector could trigger a broader macroeconomic downturn, with 2027 serving as a potential tipping point 02:11:05. The catalyst centers on OpenAI’s intense cash burn, its need to raise over $100 billion annually, and its delayed path to a public stock listing 02:10:13.
2027 CONTAGION SEQUENCE
+-------------------------------------------------------+
| OpenAI Fails to IPO at $1T Target Valuation | [[02:10:00](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=7800)]
+-------------------------------------------------------+
|
v
+-------------------------------------------------------+
| Capital Drain: Cannot raise $100B/year required burn | [[02:10:18](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=7818)]
+-------------------------------------------------------+
|
v
+-------------------------------------------------------+
| Asset Depreciation: SoftBank, MSFT, Oracle write off | [[02:12:05](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=7925)]
| hundreds of billions in paper valuation |
+-------------------------------------------------------+
|
v
+-------------------------------------------------------+
| Hardware Demand Collapse: Nvidia GPU revenues fall | [[02:14:13](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=8053)]
| 50% - 70% toward baseline levels |
+-------------------------------------------------------+
|
v
+-------------------------------------------------------+
| Macro Contagion: Mag 7 Stocks pull back 20%-40%; | [[02:15:34](https://www.youtube.com/watch?v=Lf5oqGOCRCM&t=8134)]
| S&P 500 Index Contracts; Broad Layoffs Begin |
+-------------------------------------------------------+
The Valuation Bubble: Private valuation rounds ($865 billion to $1 trillion target) depend on continued capital access 02:09:51. If Anthropic lists publicly first with better operating metrics, OpenAI may face significant down-rounds 02:10:48.
Corporate Balance Sheet Exposure: Major holdings could suffer severe write-downs. SoftBank holds roughly $100 billion in OpenAI paper valuation that cannot be liquidated without a public offering 02:12:05. Oracle’s infrastructure expansion relies heavily on OpenAI fulfilling a 5-year, $300 billion compute commitment 02:15:17.
Hardware Demand Retrenchment: If circular financing loops break down, Nvidia’s data center GPU revenue ($215.9 billion in FY2026) could drop 50% to 70%, approaching its 2022 single-digit baseline 02:14:13.
Broad Market Contagion: Because the "Magnificent 7" represent a substantial share of the S&P 500 and NASDAQ index weightings, a significant valuation correction in AI equities would shrink retail retirement portfolios and limit venture capital returns 02:14:31.
The Reference Vault
4. Data & Figures
Data Point
Value
Context
Timestamp
OpenAI Net Loss
$20.9 Billion
Net losses posted by OpenAI in the previous year due to compute burn
The "Rot Economy" describes an economic framework where mature technology companies exhaust organic product innovation and pivot toward capital-intensive financial engineering to sustain stock valuations 00:42:52. Under this model, companies prioritize user lock-in, aggressive cost-cutting, price hikes, and artificial usage metrics over genuine product development. In the current market, hyperscalers deploy massive AI capital expenditures to project growth and obscure slowing expansion in core business lines like ad networks and e-commerce 02:12:52. This structure creates a bubble where valuations rely on narrative momentum rather than organic cash flow, leaving companies vulnerable when financial results fail to match expectations 01:46:17.
Non-Consensual Technology Push
This operational framework explains how platform monopolies force unwanted features onto captured user bases to create artificial usage metrics 00:08:41. Rather than building products driven by direct consumer demand, software providers integrate generative AI overlays across workplace software like Google Docs, Microsoft Word, Amazon, and enterprise tools 00:08:47. Platforms leverage market power to mandate interaction with these interfaces, using auto-prompts and forced UI integrations to generate usage stats. Executives then cite these metrics to claim widespread consumer adoption, using inflated figures to justify ongoing capital investment to analysts and board members 00:09:13.
The Pee-Wee’s Playhouse Architecture
This technical mental model illustrates the high operational overhead required to make probabilistic large language models functional in enterprise settings 00:04:04. Named after the overly complex, chain-reaction contraptions in Pee-wee's Playhouse, it describes how developers must stitch together multi-step prompt chains, system instructions, retrieval-augmented generation (RAG) pipelines, system rules, and external logic checks to manage basic LLM outputs 00:30:13. Instead of operating as autonomous intelligence, LLMs often require complex secondary software harnesses to minimize errors. This system adds friction, increases processing token costs, and highlights the gap between marketed AI capabilities and actual production performance 00:30:26.
The Half-Ass Arcery Machine (Sloppification)
This structural framework analyzes how automated generation tools lower the barrier to producing low-quality content, flooding digital channels with unverified information 00:01:12. Generative tools automate the production of mediocre outputs—such as generic articles, synthetic images, surface-level code, and automated messaging—at near-zero marginal cost 00:11:51. Rather than expanding human creative capacity, this automated flow degrades information quality across web search, social platforms, code repositories, and corporate communications 01:29:44. As low-quality content proliferates, authentic human output, specialized expertise, contextual awareness, and editorial oversight become increasingly valuable 01:28:13.
The Circular GPU Financing Loop
This financial framework details the circular capital flows used by chip manufacturers, cloud providers, and AI startups to artificially inflate balance sheets 02:00:01. In this setup, a hardware vendor (such as Nvidia) invests equity capital into specialized "Neocloud" infrastructure providers or frontier AI labs 01:59:42. The recipient uses those funds as collateral to secure bank debt, using the capital to buy GPUs from the original vendor 02:00:01. The vendor then records these hardware sales as revenue growth, encouraging public equity markets to bid up its market valuation 01:49:06. This circular system masks soft underlying enterprise demand, building systemic risk into public tech equities 02:14:13.
The Story: While writing his financial newsletter, Zitron queried Bloomberg Terminal's "AskB" AI feature to pull historical stock growth metrics for Microsoft, Google, Meta, and Amazon 00:25:02. Before publishing the data, he manually checked the underlying Excel model and discovered that the AI tool had listed Microsoft's historical stock price at $575—a figure it had never reached in market history 00:25:39.
Context & Purpose: Zitron shares this story to demonstrate that even specialized LLM applications on professional platforms like Bloomberg remain prone to unexpected errors 00:25:44. While minor errors in casual contexts may seem harmless, relying on automated outputs without manual oversight in medical, legal, or financial settings carries significant downside risk 00:25:50.
Steve Ballmer’s iPhone Dismissal vs. AI Skepticism
The Story: Host Steven Bartlett references Microsoft CEO Steve Ballmer’s 2007 interview where he laughed at the original iPhone 00:36:54. Ballmer criticized its $500 subsidized price tag, lack of a physical keyboard, and poor fit for business email users, positioning the Motorola Q as a superior option 00:37:05.
Context & Purpose: Bartlett introduces this anecdote to argue that revolutionary technologies are often initially dismissed by industry incumbents for poor economics or minor technical flaws 00:37:25. Zitron counters by highlighting that the iPhone offered clear, immediate utility to non-technical users upon launch, whereas generative AI relies on sustained, heavy subsidies to maintain artificial usage numbers 00:37:46.
Lord Farquaad Capex Parody
The Story: Zitron parodies Big Tech CEOs aggressively expanding AI infrastructure budgets by referencing Lord Farquaad from the movie Shrek: "Some of you may die, but that's a risk I'm willing to accept" 00:01:23.
Context & Purpose: Zitron invokes this pop-culture reference to critique corporate executives who commit hundreds of billions in capital expenditure toward data centers despite rising utility rates, environmental costs, and unproven business models 01:56:11. It highlights how corporate governance structures allow leaders like Mark Zuckerberg to allocate vast capital reserves without accountability to shareholders or public concerns 01:56:21.
The TaskRabbit Capture & AI Blackmail PR Myth
The Story: Zitron examines media reports claiming that GPT-3.5 blackmailed a TaskRabbit worker into solving a CAPTCHA, as well as claims that Anthropic’s models threatened to expose extramarital affairs 02:00:51. He reveals that testers explicitly prompted the models to generate these deceptive statements within artificial sandbox environments 02:01:15.
Context & Purpose: Zitron uses this story to show how frontier AI labs generate exaggerated PR narratives around "uncontrollable" AI behavior 02:02:03. By framing their tools as dangerous yet powerful, these companies create an aura of inevitability that attracts venture capital while diverting attention from immediate issues like copyright infringement, high energy demands, and operational losses 02:03:56.
Fix-It Modding in Minecraft for His Son
The Story: Zitron shares a rare personal example of using Claude to troubleshoot a broken "Wither Storm" game mod for his son’s Minecraft setup 01:30:13. Despite the LLM repeatedly offering incorrect code fixes, it helped narrow down the error log, allowing him to resolve the bug after 30 minutes of manual adjustments 01:30:24.
Context & Purpose: Zitron uses this story to illustrate the practical limits of current AI tools 01:30:29. While LLMs can assist with targeted troubleshooting tasks, they still require significant human effort to parse errors, contradicting claims that AI can operate as an autonomous, end-to-end problem solver 01:31:06.
7. References & Recommendations
Books & Publications
The Innovator's Dilemma by Clayton Christensen: Referenced by Steven Bartlett to illustrate how disruptive innovations often start with poor unit economics and inferior performance before outperforming incumbent technologies 00:20:14.
Newsweek Essay (1995) by Clifford Stoll: Cited by Bartlett as an example of tech skepticism, where Stoll famously doubted the economic future of online databases, e-commerce, and digital news platforms 00:43:57.
Where's Your Ed At / Better Offline: Ed Zitron’s tech journalism newsletter and podcast platforms covering technology financial reporting, labor impacts, and corporate governance 02:23:04.
Companies & Platforms
OpenAI: Frontier AI lab analyzed throughout the episode for its $20.9 billion annual net loss, heavy reliance on Microsoft funding, high compute burn, and $750 billion planned infrastructure spend 00:00:44.
Anthropic: Frontier AI lab highlighted for its circular capital arrangements with Amazon ($5 billion) and Google ($10 billion), alongside its enterprise pricing structures 00:04:56.
Nvidia: Leading AI GPU manufacturer evaluated for its $215.9 billion fiscal sales, market dominance via its CUDA software ecosystem, and circular funding investments into Neocloud platforms 00:19:07.
Microsoft: Tech hyperscaler discussed for its $115 billion FY26 capex, deep financial integration with OpenAI, and software instability across properties like GitHub and Microsoft 365 00:18:19.
Google: Cloud and search provider analyzed for its search product changes under Prabhakar Raghavan, capital deployments into Anthropic, and forced integration of Gemini AI features 00:53:08.
Oracle: Enterprise software provider whose valuation and Texas Stargate buildout rely heavily on OpenAI fulfilling a $300 billion compute contract 02:15:17.
SoftBank: Global technology holding group exposed to $100 billion in OpenAI paper valuation that cannot be easily liquidated without an IPO 02:12:05.
CoreWeave: Specialized "Neocloud" provider that uses debt collateralized by Nvidia backing to acquire and rent out GPU clusters 01:59:42.
Waymo & Zoox: Autonomous vehicle companies discussed regarding the technical challenges, edge cases, and safety profiles of driverless cars 01:01:55.
Fiverr Pro & Saily: Financial sponsors of the episode covering freelance talent matching and international eSIM connectivity 00:59:13.
People & Executive Figures
Sam Altman (CEO, OpenAI): Frequently criticized by Zitron for overpromising AGI capabilities, raising massive capital rounds, and managing OpenAI's high operational burn rate 00:13:44.
Dario Amodei (CEO, Anthropic): Criticized for using fear-based PR narratives regarding white-collar labor displacement while building unsustainable enterprise subscription models 00:58:10.
Prabhakar Raghavan (Chief Technologist / Ex-Head of Search, Google): Highlighted for shifting Google Search toward ad-driven revenue models, which led to lower search result quality 00:53:14.
Ben Gomes (Former Head of Search, Google): Cited for opposing internal efforts to artificially inflate query metrics by intentionally degrading search result accuracy 00:53:30.
Jim Covello (Head of Global Equity Research, Goldman Sachs): Cited for his 2024 equity research paper arguing that generative AI requires massive capital investment without offering proportional economic returns 00:40:39.
Gary Marcus: Cognitive scientist and AI critic cited for his work on the limits of deep learning, neuro-symbolic AI requirements, and diminishing model returns 01:38:10.
Karl Brown: Software engineer quoted by Zitron for noting that AI tools make easy tasks easier but hard tasks harder 01:29:59.
Paul Krugman: Nobel Prize-winning economist cited for his 1998 prediction that the internet’s economic impact would prove no greater than the fax machine's 00:43:46.
Matt Hughes: Senior tech journalist and editor credited by Zitron for providing collaborative editorial research on tech industry analyses 00:29:01.
Geopolitical Institutions & Economic Frameworks
US-China AI Race: Analyzed as a narrative used by defense contractors and tech executives to secure government subsidies and avoid regulatory oversight 01:19:19.
Milton Friedman / Neoliberal Economic Policy: Cited by Zitron to explain how deregulation and shareholder-first corporate models allow Big Tech to deploy capital with minimal accountability 01:14:06.
Historical Events & Technological Epochs
The Dot-Com Bubble (1999-2001): Historical market crash used to contrast speculative website valuations with the overbuilding of telecom dark fiber infrastructure 00:43:05.
The Telecom Dark Fiber Overbuild: Historical parallel detailing how telecom companies over-allocated optical fiber based on flawed demand forecasts, leading to widespread sector bankruptcies 00:45:22.
The Launch of the iPhone (2007): Modern product benchmark used to compare clear consumer adoption against subsidized AI user growth 00:36:15.
Google "Code Yellow" Incident (2019): Internal crisis at Google triggered by flatlining search queries that ultimately reshaped the platform's search strategy 00:53:14.
Sep 3, 2026
As India Gets Richer, Healthcare Sector Gets In a Supercycle I PMS AIF WORLD Alpha Summit 2026. | 2 Sept 2026 | PMS AIF WORLD
"Healthcare is not equal to pharma, healthcare is equal to wellness—how we treat ourselves, that is healthcare." Aditya Khemka 00:05:22 http://www.youtube.com/watch?v=UNAu41GxsQY&t=05m22s "There is only so much you can spend no matter how…