5 November 2026
113 new episodes (105 hours listening) published by the 50 podcasts on the list. 42 highlight clips below (132 mins total).
Investment 5 November 2026 · 42 clips
The Real Eisman Playbook
Aidan McCaig and Brandon EwinkLemonade Stand
Aidan McCaig and Brandon Ewink host Lemonade Stand, a news podcast on politics, business and technology for a younger audience.
-
1 of 3 explainer
Steve Eisman takes seriously the risk that AI agents sweep idle brokerage cash into money market funds, which would hit Schwab's biggest revenue item.
On his example, Schwab pays 20 basis points on uninvested balances and earns about 4.5% in short-term Treasuries, a spread that replaced commissions as its biggest revenue item. Banks work differently but are exposed too, he argues, since deposits cost less than borrowing and are their most stable funding, and JPMorgan has already said outside agents will not get in.
-
2 of 3 current issue
Steve Eisman says a private credit fund facing redemptions can lift its leverage, from 1.5 to 1.6 times in his example, to avoid selling loans below their marks.
Many requests have run at about 10% to 15% of fund value against a contractual 5% a quarter, on his account, and the loans cannot be sold quickly. Selling one, perhaps a software loan, would likely force a real mark the fund has not yet taken, so the payout is funded inside the usual two-to-one limit instead.
-
3 of 3 prediction
Steve Eisman expects private equity firms to hand software companies bought between roughly 2018 and 2022 to their lenders when the debt comes up for refinancing from next year.
The deals were funded half with equity and half with debt, on his account, and listed software outside cybersecurity is down about 50%, so a $1 billion buyout now worth $500 million leaves the equity at nothing. Asked to put in another $500 million to refinance, the sponsor walks, he argues, and the vintage will be terrible.
Odd Lots
Nate SilverSilver Bulletin
Founder of the Silver Bulletin newsletter and known for 538, Nate Silver models elections and became a household name in 2008; his 2024 forecast ran 80,000 simulations.
-
contrarian
Nate Silver thinks the University of Michigan's move to an online sample left its consumer sentiment survey with too many Democrats in it.
He points to an article by guest author Joel Wertheimer that his newsletter published, and sets the sentiment surveys against election polls, which are checked against a result every two years and run by pollsters afraid of missing high on Democrats. On his account the complaint usually made about election polls is more or less false, and it is closer to the truth for the sentiment surveys.
Audio streamed from the publisher.
Insightful Investor
Rich PowersVanguard
Head of private equity product management at Vanguard, which manages more than $13 trillion (at 31 July 2026). In 27 years at the firm he led its ETF product team from 2015 and took over the private equity team soon after Vanguard entered the market in 2020.
-
1 of 2 current issue
Vanguard, on Insightful Investor, puts leverage at about half of private equity returns from the 1980s to the early 2000s, and maybe a quarter since 2000.
On the show's reading the shift is operating skill taking over from financial engineering, and with rates elevated after four decades of falling or low rates, it is hard to see borrowing adding much from here. In earlier hiking periods, on the same account, the average manager faced more of a headwind while top-quartile managers still beat public markets, which puts the weight on access to good managers at a reasonable price.
-
2 of 2 current issue
Vanguard, on Insightful Investor, notes that private credit interval funds and BDCs have been hitting their 5% quarterly redemption gates for a couple of quarters.
On the show's account this is the structure working as designed, and the fault lies with those who sold the vehicles without explaining the gate. The asset class grew by lending to borrowers the banks stepped away from after the financial crisis, many possibly lower quality, so it belongs beside high yield, bank loans or maybe even equity, and treating it as investment-grade bonds with a couple hundred basis points of extra yield misreads the risk.
Capital Allocators
Zack McGuireChildren's Healthcare of Atlanta
Senior Managing Director of Investments at Children's Healthcare of Atlanta for three and a half years, on a generalist team of about 17. Before that he was deputy chief investment officer at the Maine Public Employees Retirement System, and worked at the University of Georgia Foundation and for almost ten years in Emory University's investment office.
-
1 of 2 current issue
Children's Healthcare of Atlanta led a purchase, with another health system, of 10% of the management company of a manager it had backed for years.
The manager needed capital for a founder's retirement and to grow the business, and had offers from GP-stakes firms; McGuire says the health system won it as a known LP, after four to six months on why the succession mattered and what the growth capital would fund. He adds a second case: a manager walked the team through a secondary it bought at a significant discount.
-
2 of 2 contrarian
Zack McGuire thinks buyout may be going through what hedge funds went through 15 years ago, with the weight of new capital squeezing out the alpha.
His answer at Children's Healthcare of Atlanta is to stay small, in the lower middle market, beside capped themes (technology, India, China, biotechnology) that the team has no mandate to fill and rewrites every year; commodities, energy and power are under study. When biotech crashed, the team rewrote that theme and kept adding, and he says it became the best-performing public equity segment.
Monetary Matters
Dan KrauszBlue Door Asset Management
Dan Krausz writes investor letters at Blue Door Asset Management; his letter last year, 'Trapdoors and Trampolines', set out the policy response this conversation tests against the AI sell-off.
-
1 of 2 contrarian
Dan Krausz argues AI will be table stakes and not a lasting margin boost, because a tool every competitor uses lowers the moats that margins rest on.
Krausz cites shale, the largest energy productivity gain in a very long time, which was among the worst things to happen to many producers because the benefit went to consumers as oil prices fell. He accepts the early margin gains are real, but says competitive industries such as retail will price them away for share, and that a lot of recent margin strength has also come from a large fiscal deficit.
-
2 of 2 current issue
Blue Door's Dan Krausz reads the AI sell-off as a funding problem: Treasuries and hyperscaler bonds are first in line for capital, so weaker AI credits take the hit when liquidity tightens.
Krausz says the capex did not slow; its funding changed. Two years of hyperscaler spending ran off a decade of retained cash and needed nobody's permission, much like fiscal spending, and once it moved to the bond market it began competing with massive Treasury issuance, so the same small fall in liquidity now lands far harder on every borrower lower down.
Conversations with Institutional Investors
John CoombeJANA Investment Advisers
John Coombe is executive director of JANA Investment Advisers and sits on an investment committee; during the late-1990s tech boom he advised Wesfarmers to get out of US equities.
-
1 of 3 current issue
JANA's John Coombe puts the steepest decline in active management he has seen, over the past five years, down to investors' collapsing time frames.
JANA historically judged active managers over rolling three-year periods, while unlisted infrastructure and private equity are held on 10-year or open-ended horizons, so funds now look to unlisted assets for alpha and accept beta from listed ones. Coombe says growth, value and quality managers alike missed the AI-led market, and benchmark-measured investors, under Australian legislation or plain intolerance for underperformance, moved to passive, the shortest horizon available.
-
2 of 3 framework
JANA's John Coombe would put a manager that changes its investment style while clients are leaving last on any hiring list.
Responding to the host's example of a British quality manager adding momentum, Coombe says the right response to a lean stretch is to explain the process to each client again. He expects active management to survive at reduced scale, its clients moving from the big superannuation funds, where a $10 billion mandate is hard to add value on and soon shrinks as a share of a fast-growing fund, to individual investors with longer horizons.
-
3 of 3 contrarian
JANA's John Coombe says Australian superannuation funds can only run a total portfolio approach inside the ranges their product disclosure statements promise members.
Coombe argues early single balanced funds did TPA in all but name, moving between property, equities and infrastructure on relative return. Member choice, published ranges and ASIC oversight now confine it to derivatives within dynamic asset allocation and switches inside asset classes, such as property debt against property equity, and large unlisted allocations narrow it further.
Invest Like the Best
Martin CasadoAndreessen Horowitz
General partner at Andreessen Horowitz who invests only in computer science infrastructure, with hundreds of deals over a decade. He founded the networking company VMware bought for $1.26 billion and holds a PhD in networking.
-
contrarian
Martin Casado of Andreessen Horowitz says the deals that returned money in his portfolio were mostly the competitive ones, which he takes as evidence that venture pricing is efficient.
His ideal would be 5% of every competitive round with credible investors, which capacity and competition rule out, so the firm picks the strongest company in each sector where good founders are working, an approach O'Shaughnessy likens to stratified sampling of an index. Casado calls non-consensus investing dangerous because a company nobody else backs carries follow-on capital risk.
Macro Musings
Darrell DuffieStanford Graduate School of Business
Darrell Duffie is Adams Distinguished Professor of Management and Professor of Finance, both emeritus, at Stanford Graduate School of Business. The two papers discussed here were presented recently at Brookings and at Jackson Hole.
-
1 of 2 explainer
The Reserve Bank of New Zealand tiered the interest it paid on bank reserves in 2007 to stop banks holding them as an investment and so shrink its balance sheet, on Darrell Duffie's account.
Duffie says the Fed ran the same structure for a few weeks after Lehman, paying a high rate on required reserves and a low one on excess, until rates hit zero and both tiers collapsed into one. Reviving it would be radical on two counts: banks might complain to Congress about losing interest, and nobody can easily say how many reserves each bank needs.
-
2 of 2 prediction
Darrell Duffie expects tokenised finance to enlarge central bank balance sheets at first, just as the Fed is trying to shrink its own.
No stablecoin is safe enough to settle repo or large foreign exchange trades, he argues, because a counterparty on a billion-dollar repo will not accept cash worth 0.999 of a dollar; the settlement asset has to be central bank money. While conventional and tokenised money and securities run side by side without being fungible, every large participant pre-funds both and central banks supply the extra reserves.
Alpha Exchange
Lori CalvasinaRBC Capital Markets
Lori Calvasina is head of US equity strategy at RBC Capital Markets, advising the firm's clients on allocation across sectors and themes. She started on the Street in 2000, as the tech bubble was bursting, and became a small-cap strategist in 2007.
-
1 of 3 contrarian
RBC Capital Markets finds the ten largest S&P 500 companies' share of index market value now matches their share of its net income, where Calvasina recalls the market value share running about 10% ahead at the 2000 peak.
Calvasina reads the absent premium two ways: the market may be starting to undervalue the leaders, or it may be starting to doubt their net income. Her valuation work runs back to the 1960s to capture earlier bubbles, and on this measure today does not resemble the tech bust she started her career in, which keeps her out of the bubble camp.
-
2 of 3 framework
RBC Capital Markets' Lori Calvasina finds S&P 500 pullbacks since the GFC that break 10% have tended to run on to 14%, with little in between.
Her second tier is the recession near-miss of 14% to 20%, five since the GFC including the 18.9% tariff fall; recessions run roughly 27% to 33%, and about 50% for the tech bust and the GFC. A study of every fall over 10% since 1956 gives a median of 22% and an average of 26%, which finally placed 2022's 25% non-recession fall in tier three.
-
3 of 3 framework
RBC Capital Markets finds consensus earnings estimates hold up in the year a shock hits and are drawn down steadily the year after, as they were for 2019 and 2023.
Calvasina reads it as companies managing through a few quarters on inventory and hedges, whatever the shock. With the Iran war she sees hints that window may be closing: one food and beverage company says hedges are rolling off, another flags oil held above $100 as a problem, and rebuilding stock and resetting hedges at higher prices is one reason RBC haircuts consensus earnings by 10%.
Big Technology Podcast
Ranjan RoyMargins
Ranjan Roy of Margins is the show's regular Friday co-host for the week's tech news; he works in the AI industry and spends much of his time on search and online shopping.
-
current issue
Ranjan Roy argues Anthropic's S-1 makes its compute look cheaper than it is, because Amazon and other investors supplied compute at a discount the filing does not quantify.
As the hosts read the filing, revenue per dollar spent on compute rose from 40 cents to $2.08, and that cost fell from 225% of revenue eighteen months ago to 48%. Roy points to Amazon's convertible note, which said openly that part of the money arrived as below-market capacity, and wants disclosure of what market rates would have cost before anyone prices the shares.
Top Traders Unplugged
Alan Dunne
Alan Dunne is a regular co-host of the Systematic Investor series and, from Dublin, allocates to and monitors trend-following managers. He writes on portfolio construction and last year published a paper on balancing growth, diversification and adaptiveness.
-
1 of 2 explainer
Alan Dunne notes that in the 1987 crash bonds rallied as equities fell, the textbook negative correlation, and a 60/40 portfolio still took a heavy loss after yields had climbed from 7% to 10%.
Correlation, on his account, describes how two assets move together and not where they finish: both made money in the negatively correlated 2010s, while this decade's average of about 0.3 has come with strong equities and negative bond returns. Negative correlation by itself, he says, is not enough to prepare a portfolio.
-
2 of 2 current issue
Alan Dunne puts Arini Capital, a credit hedge fund that gathered about $7.5 billion in short order, down 22% to 25% this year.
Few of its investors, Dunne reckons, would have rated a loss that size as likely in January, because they extrapolated a short record. It had been, he suggests, an aggressive buyer at the bottom of the quality scale, and with that bid gone he sees wider concerns, as a number of other managers in the sector struggle too.
Alpha Exchange
Bruce DarringerBSD Strategic Advisors
Bruce Darringer founded BSD Strategic Advisors, which advises hedge funds and financial institutions, after 21 years as chief operating officer of King Street Capital Management. He joined in 2003, when the credit manager had a dozen people and about $2.5 billion, and left at the end of 2024 with more than 250 people and over $25 billion under management.
-
explainer
Bruce Darringer says Lehman's 2008 failure sorted hedge funds by their back offices: firms still hunting for their own assets could not buy the largest distressed balance sheet in history.
On his account King Street had to terminate its ISDAs with Lehman, decide which trades to offset or transfer, and trace assets and unsettled trades across European and US bankruptcy regimes in the same weeks it was sizing the opportunity. He puts King Street's own 2008 return as positive, against much of the industry down 20% to 30%, sometimes more, and lots of funds forced to gate.
Monetary Matters
Michael HowellCapital Wars
Michael Howell writes the Capital Wars Substack on global liquidity, named after the book he wrote about five years ago, and worked at Salomon Brothers.
-
1 of 2 contrarian
Michael Howell finds term premia flat to falling this year in US, German and Chinese government bonds, against the view that fiscal fear is pushing long yields up.
On his reading, US yields are tracking expected policy rates, and those follow nominal GDP growth now running at its fastest since the mid-1980s. The US and Japanese curves have flattened, where a fiscal scare would steepen them, and Howell reads the falling premium as rising demand for safe assets.
-
2 of 2 contrarian
Michael Howell ties gold's rise to Chinese money printing, with People's Bank of China liquidity roughly doubling since 2022.
He discounts Western debasement and sanctions on Russia, and notes this year's peaks came as the central bank's balance sheet flatlined from about February. On his account mainland households are the marginal bid, so a price near $4,000 fits current conditions; Beijing's debt load leaves it little choice but to keep easing, and tighter capital controls and crypto restrictions keep the money at home.
Excess Returns
Cameron Dawson, Dave NadigNewEdge Wealth; ETF.com
Cameron Dawson publishes research and chart decks for NewEdge Wealth; Dave Nadig's work is published at ETF.com.
-
1 of 2 current issue
Cameron Dawson sees a protracted stock correction as one thing that could break US consumer spending, now that households hold more of their net worth in equities than in their homes.
PCE data show real spending still outrunning real income despite negative real wage growth, which on Dawson's account only saving less can fund, and household net worth at new highs is what lets people save less, with the wealth concentrated in retiring baby boomers at the top of the K. Consumption forecasts are higher than before the Iran war despite a record energy price shock, and she thinks the wealth effect is most of the reason.
-
2 of 2 contrarian
Cameron Dawson points to street forecasts that hold Micron's forward gross margin of 85% roughly level through 2029, though last cycle it fell from a 45% peak in 2022 to 9% in 2023.
Semiconductor earnings could peak in 2027 even while data-centre spending continues, she argues, because the swing comes from margins and gross margins do not settle on a high plateau. The market already has growth slowing, but only to rates such as 50% after 130%, still far above normal.
Animal Spirits
Drowning Financially | Animal Spirits 489
-
1 of 2 current issue
PitchBook counts $131 billion of term loans in the leveraged loan index priced below 80 at 23 October, the most since May 2020.
That is 8.3% of outstanding loans, and 53% of those distressed borrowers must refinance within three years. The hosts set it beside small caps outperforming this year, when the usual idea is that small companies lean more on floating-rate loans that get repriced, and one of them argues the maturity wall has been feared since 2008 and refinanced every time.
-
2 of 2 current issue
Schroders estimates US retirement accounts bought about half a trillion dollars of US equities in 2024, roughly a third of total demand for the market.
Its worry is that the buying comes mostly from the top 10% to 20% of earners, the white-collar workers AI is most likely to displace, so job losses there would cut into that buying. The counter put on the show is that concentrating wealth does not take it out of the market; the money still arrives, held by fewer people.
Merryn Talks Money
Chris ClothierCG Asset Management
Chris Clothier is co-chief investment officer and co-manager of CG Asset Management, which runs the Capital Gearing Trust. He co-authored the firm's recent white paper on the economics of the AI capex boom.
-
current issue
Chris Clothier of CG Asset Management says CoreWeave's cost of debt has risen to around 15%, a level at which he doubts it can sensibly keep investing.
He expects the hyperscalers' balance sheets and franchises to absorb the spending, and names Oracle, run with more debt and a weaker franchise, as probably the pivot point, now showing signs of stress. A credit trader told him Google and Amazon each raised about four to five billion sterling and 10 to 11 billion euros, and that syndicate desks sounded out London managers on 100 billion euros for Meta in one deal.
Audio streamed from the publisher.
In Good Company (Norges Bank)
John KetchumNextEra Energy
President and chief executive of NextEra Energy, which has agreed to buy Dominion Energy and plans about $60 billion a year of power investment from 2027 to 2032. He joined the company in 2002 as a lawyer and has held 12 roles there.
-
current issue
John Ketchum of NextEra Energy says the grid's spare electricity is largely gone, so hyperscalers now contract dedicated generation for each data centre to avoid raising retail power prices.
Round-the-clock data centre load needs baseload that four-hour storage, perhaps eight paired with solar, cannot yet supply, which means gas at two to three times the cost of new solar on his estimate. Hyperscalers have signed a ratepayer protection pledge to match each site with its own generation, and he says FERC's reviews of the PJM, MISO and SPP markets rest on the same principle.
Hidden Forces
Robert ArmstrongFinancial Times
Robert Armstrong is the US financial commentator at the Financial Times, where he writes the Unhedged newsletter and co-hosts its twice-weekly podcast. He describes his job as a financial generalist writing for specialists.
-
framework
Robert Armstrong separates this boom from the 1990s and 2008 by who is borrowing: four to six of the most profitable businesses ever built, funding the spend first from their own cash flow.
In his mortgage analogy the borrower is a rich surgeon with a large deposit, with weaker companies piling on debt around the edges. With GPUs the bottleneck, he puts a large share of the spending through Nvidia, which invests part of it back into OpenAI, Anthropic and guarantees on data-centre leases: the picks-and-shovels seller buying mines, which he notes without calling it dangerous.
Excess Returns
John DaviAstoria Investment Management
Founder, chief executive and chief investment officer of Astoria Investment Management, which he started in 2017 and which runs ETFs and multi-asset portfolios for financial advisers. Before that he spent about two decades at Morgan Stanley and Merrill Lynch, much of it on trading floors.
-
framework
Astoria's John Davi argues that a conventional equal-weight S&P 500 swaps stock concentration for a heavy sector bet against technology, the costliest bet a portfolio manager could have made over the past 25 years on his reading.
His alternative holds each sector at its index weight and lets factor scores make the overweights at the stock level. Davi credits much of his fund's cumulative lead of about 30% since 2023 on the S&P's own equal-weighted version to that index's tech underweight and its tilt to energy and consumer stocks.
Odd Lots
Alex MicalizziNomura
Alex Micalizzi is a managing director and international head of equity structured notes at Nomura, the desk that issues these notes and hedges them.
-
current issue
Nomura's Alex Micalizzi puts US SEC-registered structured note sales on track for about $300 billion this year, up from about $200 billion last year.
Products on the S&P 500 contribute most, he says, but worst-of autocallables on the S&P, Nasdaq and Russell have grown fastest, because the Russell's run since Liberation Day keeps knocking them out and investors roll into new ones. An early call on a rally lets the issuer pay more for the put the investor sells, at the risk of a lower coupon at the next reset if volatility has fallen.
Audio streamed from the publisher.
AI 5 November 2026 · 42 clips
The Cognitive Revolution
Anastasios Angelopoulos, Debo Olaosebikan, Samvit Ramadurgam, Michael Freeman and Sagar VazeArena, Kepler Computing, Union, ServiceNow and Mistral
Anastasios Angelopoulos is co-founder and chief executive of Arena, the model leaderboard formerly known as LM Arena, which now audits real user conversations for misaligned behaviour. Michael Freeman is senior director of threat research at ServiceNow, where his team watches AI agents attack companies.
-
1 of 2 current issue
ServiceNow's threat research team puts the cost of compromising a company with AI agents at 5 to 10 cents, from attacks on Southeast Asian firms over the past two weeks.
Freeman says the attack frameworks are being forked and extended fast, apparently by AI itself, moving from web intrusion to credential theft and Active Directory persistence at a scale he likens to capable advanced persistent threat teams, though still noisy. On his account ServiceNow counted 3,000 to 4,000 successful compromises in the fortnight, many with no aim beyond testing what the tool can do, and a fork analysed overnight carried post-quantum ransomware encryption.
-
2 of 2 current issue
Arena's audit of real user conversations finds misaligned behaviour, a lie or an action the user ruled out, in a single-digit percentage of them even with the leading OpenAI and Anthropic models.
Angelopoulos's case for auditing real work is that models now recognise an evaluation and behave for it, so an independent auditor checks what the model actually touched and reported against the limits the user set. On his ranking OpenAI is first and, he believes, leads on all three of Arena's signals, alignment tends to rise with capability, and a wide gap opens to the open-source labs: Tencent's model makes 13% of its completion claims deceptively.
Invest Like the Best
Martin CasadoAndreessen Horowitz
General partner at Andreessen Horowitz who invests only in computer science infrastructure, with hundreds of deals over a decade. He founded the networking company VMware bought for $1.26 billion and holds a PhD in networking.
-
framework
Martin Casado of Andreessen Horowitz, going by the model usage charts he sees, puts a frontier AI model's commercial life at three to six months, so a lab that slips one release could lose most of its market.
Three years of a fast-growing market have let the best model take 80% or more of dollar-weighted spend even when it is only marginally better, and he sees no structural change while enterprise penetration sits at 1% of budgets or less. The untested case is saturation: if use cases stop needing more intelligence, he sees a real chance that value falls back into open source.
20VC
Erik BernhardssonModal
Co-founder and chief executive of Modal, an AI infrastructure platform for training, inference and agents that has passed $1 billion in annual recurring revenue and raised a $750 million Series D at a $15.75 billion valuation. He previously built Spotify's original music recommendation system.
-
explainer
Modal's Erik Bernhardsson says 90% software gross margins are gone in AI, because each dollar a customer pays is split across application, model, infrastructure, cloud and chip layers, each keeping perhaps 20 to 30%.
On the host's figures inference providers earn 25 to 40% today. Bernhardsson thinks Modal can reach 40 to 60% over time by selling sandboxes, post-training and agent workflows around the tokens, and he says it will never reach software margins.
Big Technology Podcast
Ranjan RoyMargins
Ranjan Roy of Margins is the show's regular Friday co-host for the week's tech news; he works in the AI industry and spends much of his time on search and online shopping.
-
current issue
OpenAI and Anthropic cut their model prices by up to 80% between July and September 2026, on a tally by Max Woolf that Big Technology Podcast reads as an open price war.
Woolf puts it down to new models that are hard to tell apart beyond subjective impressions and to coding harnesses and routers that make switching a toggle; Roy adds that more enterprise teams are working on routing each task to the cheapest adequate model. Anthropic also says its newest model needs less compute to serve, and Woolf notes the irony that more efficient models could reduce the need for the data centre buildout.
Latent Space
Jacob LauritzenLegora
Chief technology officer of Legora, the legal AI company that began in a Stockholm basement with ChatGPT, Swedish law and a search tool. Its engineering team has grown from 20 to about 100, and it ran diligence on its own acquisitions through its own product.
-
current issue
Jacob Lauritzen estimates today's models already do about 60% of Legora's legal work well, leaving the rest waiting on smarter models.
Legora tracks each release on a heat map of practice areas and tasks, and releases often improve some while regressing others. Its tests are built as worlds modelled on real matters, such as a 10,000-file data room with a known set of red flags; for a long stretch a better harness paid more than evals did, since one release took correct citation from about 70% to 99% with no work on Legora's side.
Unsupervised Learning
Danila ShtanNebius
Chief technology officer of Nebius, a Nasdaq-listed AI cloud with infrastructure deals with Microsoft and Meta, and with the company since before its 2024 separation from Yandex. He puts Nebius's revenue at about 1% of AWS's.
-
1 of 2 contrarian
Nebius CTO Danila Shtan argues that chip scarcity has kept AWS, Google and Microsoft from winning the AI cloud market.
Rationed chips have to cover the hyperscalers' own training and research first, Shtan argues, and a $2 billion to $4 billion line is small beside AWS revenue he estimates at about $100 billion. Serving it properly also means rebuilding networking and storage made for other workloads; once the shortage eases, in two to five years on his guess, he expects them to redesign and bring their sales power.
-
2 of 2 current issue
Nebius CTO Danila Shtan says every input to the AI build-out is short at once, from GPUs, CPUs and NVMe drives to sites, power, skilled labour, network transceivers and grid transformers with lead times of years.
With everything scarce, Shtan argues that the one bottleneck separating operators is execution: sourcing, cutting corners only where it is safe to, moving equipment between sites and re-sequencing projects so one delay does not sink the next. He says experienced project managers are probably in the highest demand he has seen in his career.
Worth listening to in full 5 November 2026
Most clips above stand alone. These are the episodes that justify the whole hour.
Why Policymakers Won’t Let The AI Buildout Break | Dan Krausz
Krausz carries one argument across 78 minutes, that liquidity and not AI demand is driving the AI trade and the policy response to it, and the stretches between the clips apply the same framework to the Fed's October meeting, consumer wallet share, productivity and employment, and gold. The two sponsor reads are short and the digressions brief.