On August 26, Milk Road PRO goes from $25/month to $39/month, and annual goes from $250 to $399.

If you join before midnight ET on August 25, you'll keep today's price for as long as you're a member.

The platform now has 5 analysts running real-time portfolios: live positions visible, every trade published the moment it happens with live notifications, across AI, stocks and crypto.

Melvin runs one of the top-performing portfolios, launched in February and up 45% since. Some closed trades: Micron +145%, AMD +140%, Nebius +131%.

Figure Technology (FIGR), the onchain lending platform, recently scored north of 150 on the Rule of 40 (revenue growth plus profit margin). Only Nvidia and Micron ranked higher on the S&P 500.

Figure shifted from a capital-constrained lending model to a software-as-a-service platform where other mortgage banks, credit unions and fintechs can originate loans on its rails.

Traditional mortgage origination costs roughly $11,000 per loan and takes around 40 days. On Figure's system it costs under $1,000 and takes 5-7 days.

In Q1 2025, Figure wrote 65% of the volume moving through its own marketplace. Last quarter, partners wrote 65% of it, with partner count growing from 144 to 489.

Partners write ~2.6x more volume after joining Figure than they did before, as loans sell faster with more buyers bidding.

The Milk Road Show podcast episode: "Bitcoin Just Pumped… But Is the Bear Market Really Over?" featuring discussion on recent Bitcoin price action and market sentiment.

The real U.S. debt crisis trigger is Americans losing confidence in the ability to manage government finances and obligations.

Treasury said this morning that it would put more muscle behind an effort to steady the long-term bond market — a surprise announcement that immediately drove borrowing costs down.

The Treasury Department is increasing its purchases of long-term government debt — a mechanism that should boost demand for bonds at a time when investors have been selling them (and, in turn, pushing yields higher).

Treasury will raise the maximum size of each purchase to at least $4 billion, from $2 billion now, focusing on longer-term Treasuries in the 10- to 30-year range — exactly where yields have been under the most pressure.

The expansion will run from Sept. 9 through Nov. 4, when Treasury conducts its next regular review of its borrowing plans.

The yield on the 30-year Treasury bond fell as much as 0.1 percentage point following the announcement, an unusually sharp move so quickly. But it barely dents the run-up of recent weeks. Long-term borrowing costs remain near multidecade highs. The 30-year yield is around 5.2% this morning, up from roughly 4.63% just before the Iran war.

Canada has three days to lock in a deal with the White House before 50% tariffs hit billions of dollars of its goods after the Trump administration granted a last-minute reprieve.

Treasury Secretary Scott Bessent has shown willingness to step into markets when prices move in an unwelcome direction.

Treasury intervened to prop up the Japanese yen earlier this month, producing an immediate market reaction that later faded. That intervention also reduced the risk that Japan would need to sell U.S. Treasuries to raise dollars to support its currency.

Bessent and Federal Reserve Chairman Kevin Warsh have pushed for the two institutions to work more closely together on areas where their policies overlap, including a potential new Fed-Treasury "accord" aimed at better coordinating their approaches to government debt and the Fed's balance sheet.

Milk Road AI is now Milk Road Stocks.

AI is still a massive part of what we cover, but our world has expanded into memory, robotics, space, energy, and the biggest opportunities across the market. So we figured our name should too.

SpaceX is targeting roughly $100B in annualized revenue by the end of 2026. Based on that, fair value is estimated at roughly $131–$244 per share.

Historically, Elon-linked companies have also traded at a 40-75% premium which could push that range to $183–$427.

A recent SemiAnalysis report projects SpaceX could reach a $305B annualized revenue run rate by the end of 2027. That number comes from one thing: compute. SpaceX is attempting to add 8.6 GW of AI infrastructure in the next 16 months. The previous record was AWS at 3.8 GW in a year.

Elon values AI compute at $30-$50 per watt per year. At 10 GW, that's $300-$500B in annual revenue from compute alone.

The first lockup tranche (on August 6th) hit the market with little visible impact. But more unlocks are coming through December which could create selling pressure and volatility.

Stocks have been behaving oddly recently, with downright dull moves for the S&P 500 index, even as companies within it swing like mad.

A measure of single-stock volatility — based on readings on each S&P 500 stock from the options market — spiked to more than 50 during July, trouncing the previous decade's average of roughly 32. It last got to 50 when the Trump tariffs spooked the market in April 2025. Before that, it only got there during the depths of the pandemic. But unlike those previous panicky episodes, the key gauge of index level volatility for the S&P 500 — not its individual constituents — barely budged this time.

"We saw the same thing in the '90s, the late '90s" Ben Bowler said of the recent divergence between volatility of individual stocks and the index itself. Bowler heads up equity derivatives research for Bank of America Global Research. "That's actually quite indicative of an asset bubble brewing in tech and AI."

It's AI. In recent years, the market has been less focused on economic developments — which tend to be broadly good, or bad, news for most stocks simultaneously. Instead, the market's view of a company as an AI winner or loser is what moves shares. That's resulting in price moves that are individual one-offs rather than market-wide, correlated swings.

A lot of big hedge funds put on a trade designed to take advantage of this volatility regime. Known as the dispersion trade, it involves betting against big swings in the S&P 500 index — using the options market — while also betting on choppiness for individual stocks.

The prevailing winds shifted in July as new Chinese AI models prompted questions about whether cheaper options could compete with offerings from Anthropic and OpenAI. Stocks associated with the AI buildout tumbled. Some of those stocks were big positions of the hedge fund Situational Awareness, which was betting on them using borrowed money, or margin. When they started to tumble, it faced margin calls it couldn't meet without unloading more shares, worsening the downturn.

At the same time, macro risk re-emerged as it became clear that the Iran war was far from done and dusted. Oil prices rose, reigniting worries about inflation. And the shaky performance of Federal Reserve chairman Kevin Warsh at his first press conference last month prompted government bond yields to climb.

The result was a big unwind of trades that had previously been working well. As hedge funds pulled in risk taking, volatility fell sharply for individual stocks.

Analysts say, despite the recent trade unwind, AI is still the most important thing for the markets. So it's likely that the swings in individual stocks will stay sharper than usual, at least compared to the index, for a while. "An AI‑driven stock market is one where we should expect higher levels of dispersion relative to history as the power of disruption can accelerate the separation between perceived winners and losers," Katrina Rodriguez, a derivatives trader at JPMorgan Private Bank, wrote in a statement to Axios. "Broader impacts will be felt only if the market questions the aggregate gains to corporate revenues from AI," Goldman Sachs analysts noted last week.

Billionaire Mark Walter, chairman of Dodgers, faces controversy surrounding life insurance arrangements and their broader implications for financial markets and corporate governance.

Spotify co-founder Daniel Ek's body-scanning startup Neko Health announced that its first US clinic will open in New York City on September 24. Customers can pay $499 to undergo "non-invasive and radiation-free" scans that Neko says capture millions of data points. Neko evaluates a person's cardiovascular health, skin, blood, vital signs, grip strength, body fat and more, providing a review of how their body is functioning within an hour. The firm said thousands of New Yorkers are already on the waitlist.

Bitcoin climbs near $65,000 in the tenth month of its drawdown as Chain check shows 8 of 12 capitulation signals firing and realized volatility has collapsed to multi-year lows, now trading below the Nasdaq. Taker buy volume has entered a historical exhaustion zone while spot demand approaches its first positive turn since February, with conviction buyers accumulating in a pattern that echoes the 2022 lows.

Funding rates sit at a 20-month high reflecting net-long positioning just as the SEC proposed "Regulation Crypto Assets," a clearer framework for certain investment contracts. The market remains tightly coiled inside a rare volatility squeeze. Institutional capital continues absorbing supply while long-term holders stack through the noise. On-chain liquidity is building underneath, setting up rotation into high-throughput settlement rails.

This is the digital asset market collapsing into an explosive apex. Officially, eight out of twelve critical capitulation signals have fired across the network. At the same time, the SEC has also issued a comprehensive regulatory framework. This double catalyst comes at a pivotal time for broader macro trading.

Legacy financial plumbing and liquidity conditions continue to be tested by restrictive Fed policy. Beneath the surface, digital resources are completely decoupling from macroeconomic stagnation. The crypto market analysis shows an unprecedented squeeze in key network metrics. Smart money is aggressively front-running the next regime change while retail sentiment is exhausted.

Sideways volatility is testing undisciplined traders across all exchanges. This structural exhaustion is seen as a generational accumulation setup. The trap door is closing on latecomers.

Join Axios' Sara Fischer in Axios Media Trends, a weekly newsletter that brings you the biggest trends reshaping media.

Why it matters: From new technology and industry IPOs, to emerging user behaviors and the policy fault lines between them, there's a lot changing the media environment.

Details: Axios Media Trends publishes weekly on Tuesdays.

Disney and ABC have filed a First Amendment suit against the FCC.

Stocks fell as soaring bond yields weighed on markets. The buy now, pay later company Klarna experienced a significant decline after cutting its outlook and announcing the departures of its CFO and CMO.

La-Z-Boy shares are trading down 17% in extended trading after the furniture maker reported an unexpected 3% decline in sales for the quarter ended July 25.

Opening arguments kicked off in a landmark child safety and privacy case against Meta over allegations that Facebook and Instagram were designed in ways that harm children.

What the plaintiffs said: "What you're going to hear in this trial is how Meta hooked children on its platforms," a plaintiff attorney told advisory jurors in federal court in Oakland. "It designed platforms so that children would keep coming back."

What Meta said: Meta's attorney said the company "strongly disagrees" with the allegations. He sought to prepare jurors for evidence the states will present at trial, seeking to contextualize what he called "loose language" in internal communications. He said Meta has created tools to help people manage their time on its platforms, and argued that measuring user engagement is a legitimate business goal.

The case details: The federal trial is being argued by California, Colorado, New Jersey and Kentucky — four of the 29 states that together sued Meta in 2023. The states have previously told the court they will seek damages of about $200 billion. A potentially bigger risk to Meta — and other social media companies watching this case — is if the court compels Meta to make changes to its algorithms and product design, which could have far more lasting consequences for the social media business model. The trial is expected to last six weeks.

Home Depot delivered its best comparable sales performance on a quarterly basis in more than two years, but the sluggish housing market continues to limit the retailer's upside.

Why it matters: Muted housing sales, limited new construction and elevated interest rates have undermined home renovation projects in recent years.

Driving the news: Home Depot nonetheless posted a 1.7% increase in comparable sales, beating expectations. The home improvement chain's average transaction was $92.50 in the period ended Aug. 2, up 2.5% for the quarter, though the number of transactions fell 1% on a comparable basis.

State of play: Home Depot is still grappling with "frozen housing conditions," CFO Richard McPhail said on the home improvement chain's earnings call. He noted that housing turnover "has never been lower as a percentage of the housing stock."

Still, small renovation projects helped keep the home improvement retailer's sales lively, both among DIYers and pros. "The worst of the cycle downtrend is likely behind us," D.A. Davidson analyst Michael Baker wrote today in a note on Home Depot.

ABC filed a lawsuit against the FCC, arguing the agency has violated its First Amendment rights by retaliating against the broadcaster for its editorial decisions.

OpenAI is unveiling its long-expected teen experience for ChatGPT that features parental controls, efforts to encourage real-world relationships and a focus on learning rather than serving up answers.

Kalshi filed to offer perpetual futures on equity indexes. "Perps" are futures-style contracts that track the price of an asset.

NFL commissioner Roger Goodell said he has "no doubt" that the league will eventually have teams located outside the U.S.

In May, an OpenAI model overturned 80 years of assumptions about how geometric points can be arranged in space, disproving a famous conjecture by Paul Erdős. In August, a swarm of autonomous Claude subagents made advances in analytic number theory that appeared to compress years of mathematical labor into a single weekend. In both cases, the AI mathematicians found underlying structure and patterns that human mathematicians had not.

Public-key encryption is effectively unbreakable because factoring large numbers is practically impossible — there is no good way for an attacker without the private key to reverse the multiplication of two large prime numbers. However, if AI discovers structures and patterns that make factorization radically more efficient, it might break internet security that relies on this assumption.

There is no sign of this happening yet, and some mathematicians are skeptical it will anytime soon. However, the problem is that if it does happen, it will happen suddenly. Unlike quantum computers where hardware improves slowly and publicly, AI might make a conceptual leap on factorization that seemingly comes out of nowhere.

Not all cryptographic defenses rely on structured math. Hash-based cryptography, for example, turns data into unstructured noise, leaving hackers no equations to reverse engineer or patterns to decode. "The point of hash functions," according to Eli Ben-Sasson, co-founder of StarkWare, "is that they have no structure."

A hash function transforms data into a string of random-looking numbers, from which it is virtually impossible to work backward to the original data. Ben-Sasson believes this makes zero-knowledge proofs running on blockchains the ultimate defense against quantum computers and god-like AIs.

Standard web security uses public-key cryptography to establish a secure channel for data exchange. If the cryptography is compromised, all that data is at risk. STARKs eliminate the channel, broadcasting instead a hash-based proof that anyone can verify. That proof can be broadcast — publicly — to establish that a private transaction is valid without revealing its contents. There is nothing to decrypt or steal — only mathematical evidence that the underlying computation is correct. In short, STARKs replace secret-sharing with zero-knowledge verification.