September 27, 2026
Bonus Content: Free AI Models Still Make Money. Here’s How.
America’s Emergency Oil Reserve Just Hit A 44 Year Low
It is 40.6% full. Here is why that reaches your grocery bill and your retirement account.
284.6M
BARRELS REMAINING
Week ending September 18, 2026. EIA Weekly Petroleum Status Report, released September 21.
The Strategic Petroleum Reserve is the country’s emergency crude stockpile. Its job is to soak up a supply shock before that shock reaches the price you pay.
Federal data now puts it at 289.7 million barrels, roughly 40.6% of its 714 million barrel authorized capacity. That is the lowest level since 1982.
The short version of how it got there:
✔ Before February 28 of this year, the reserve held roughly 415 million barrels.
✔ After the Strait of Hormuz was disrupted, a chokepoint carrying close to 20% of global oil supply, President Trump authorized a 172 million barrel release in March.
✔ That release was part of a coordinated international effort, with IEA member nations collectively committing 400 million barrels. Reported as the largest emergency stock mobilization the agency has ever run.
✔ The reserve has kept draining since. It fell another 3.7 million barrels in the most recent reported week alone.
One analyst note circulated by CNBC put it bluntly, calling this another inflation impulse and saying the country effectively has no strategic reserve left to speak of.
That’s rhetoric. 289.7 million barrels is still a real stockpile, and it sits above the statutory minimum of 252.4 million barrels set under the Energy Policy and Conservation Act. Anyone telling you the tank is empty is overselling it.
But the cushion is thinner than it has been in more than four decades, and thin cushions matter for one reason.
Energy feeds into nearly everything you buy, from groceries and freight to utilities and building materials. When oil moves and there’s less reserve on hand to blunt it, more of that move ends up on the shelf. Gasoline has been running around $4.08 a gallon in recent reporting.
Inflation doesn’t arrive as an event. It works as a slow subtraction from every dollar you’ve already put away.
A retirement account does not need a crash to lose ground. It only needs prices to keep rising faster than the account grows.
This is the kind of stretch gold has historically been held for. It promises nothing about returns. It’s savings held outside the currency and outside the paper system.
Central banks seem to think so too. The World Gold Council reported they bought a net 288.9 tonnes of gold in the second quarter of this year, up 62% from a year earlier.
The tax code allows eligible IRA, 401(k), TSP, and 403(b) savings to be diversified into physical gold and silver through a properly structured self directed IRA, generally without triggering a taxable distribution when the transfer is handled correctly.
Send me the FREE Precious Metals Retirement Guide
Inside your free guide:
✔ How energy shocks have historically fed into consumer inflation, and how quickly.
✔ How gold has behaved during past inflationary stretches.
✔ How a Gold IRA generally works, and how you may be eligible to move a portion of an existing IRA, 401(k), TSP, or 403(b) into physical metals.
✔ How physical metals can help diversify savings outside the paper system.
✔ A simple, conservative way to get started.
Or call 1-888-691-8238 to speak with a precious metals specialist.
The reserve was the cushion. There’s a lot less of it now.
Free AI Models Still Make Money. Here’s How.
Giving away your core product sounds like a losing strategy. In AI, it has become a surprisingly effective one. Understanding how requires looking past the model itself.
The Complement Play
The dominant strategy is something strategists call “commoditize your complement.” Open source becomes marketing, a feedback engine, and a way to undercut whatever layer sits above the model in the value chain. The model is free. Everything built on top of it is not.
Meta is the clearest example. Meta’s monetization engine was never Llama itself. The company has the distribution and advertising infrastructure to profit from leading-edge AI, but if it depended on a rival’s model, value would accrue to that vendor instead. By releasing Llama under a community license, Meta works to keep foundation models from becoming a tollbooth controlled by a single competitor, while protecting its core social media and ad businesses.
While “selling access” to Llama is not Meta’s business model, the company has disclosed revenue-sharing agreements with some companies that host Llama models. Companies with over 700 million monthly active users need special licensing from Meta under the Llama community license, effectively giving the company veto power over major deployments. Free, with conditions.
The Enterprise Services Layer
Mistral runs a sharper version of the same logic. Mistral makes money from a paid developer API billed per token, its Le Chat assistant on free and paid tiers, and enterprise deployments. Many of its models ship as open weights to drive adoption, then funnel serious usage into the paid API and enterprise contracts.
Reuters has reported that Mistral has discussed a $1 billion recurring revenue target by the end of 2026. The open weights are the top of the funnel. Mistral’s strongest differentiation is control, not benchmark leadership. Open weights, self-hosting, regional inference, model customization, and private deployment make the platform especially relevant to regulated and industrial organizations. Governments do not want their data routed through American servers. That anxiety is a revenue line.
Mistral’s Workflows product pushes its role from a model provider toward a workflow platform. Workflows launched in April 2026 as part of Studio, making orchestration a more explicit revenue surface, and each workflow built in Studio can increase inference consumption per enterprise customer.
What Traders Should Watch
Open source AI products typically monetize through hosted inference, deployment tooling, governance, and fine-tuning workflows rather than code access alone. Customers pay for reliability, observability, and secure enterprise usage. The model is a door. Revenue lives in the rooms behind it.
The trading lesson here is structural. When evaluating any AI company releasing models for free, the question is never whether giving away the model destroys value. It is which layer of the stack that company actually controls. Free at the foundation is only a problem if a firm has nothing above it to sell.