Sunrun and Voltus this week turned home batteries into a line item hyperscalers can buy directly — a "Bring Your Own Capacity" deal that lets AI data-center operators fund residential VPP capacity the same way Google already committed 100 MW to, and it lands on a residential segment whose federal tax credit is gone for good but whose storage attach rate hit a company record earlier this quarter regardless. Meanwhile EIA quietly cut its own Henry Hub price forecast and lithium spiked nearly 10% in a week on the other side of the world.

MACRO
EIA cuts its own gas forecast as WTI climbs and storage keeps building a cushion

WTI opened at $85.76/bbl on Aug 20, continuing a firmer tone after weeks of choppier trading (FXDailyReport). Natural gas told the opposite story: the EIA's August Short-Term Energy Outlook cut its 2026 Henry Hub forecast to $3.44/MMBtu (from $3.67 in July), citing record production and softer LNG feedgas demand, and now expects spot prices to stay under $3.00/MMBtu until November (Rigzone). The Weekly Natural Gas Storage Report backs that read: working gas hit 3,169 Bcf as of Aug 14 (+16 Bcf w/w), 185 Bcf above the 5-year average and tracking toward a record 3,985 Bcf by end-October — the biggest pre-winter cushion since 2016 (EIA/IndexBox).

Crude told a tighter story: the Weekly Petroleum Status Report put commercial inventories at 428.8M bbl (week ending Aug 14), exactly at the 5-year average, with refinery utilization climbing to 97.2% as refiners push hard into late-summer demand (EIA/IndexBox). Oil is firming on utilization and imports pulling back (-746K bbl/d w/w); gas is loosening on production outrunning demand. Two different macro tapes feeding one grid.

• WTI: $85.76/bbl (Aug 20); Henry Hub 2026 forecast cut to $3.44/MMBtu (from $3.67), EIA STEO.

• Gas storage: 3,169 Bcf (Aug 14, +16 Bcf w/w), 185 Bcf above the 5-year average, record build pace.

• Crude inventories at the 5-year average (428.8M bbl) with refinery utilization up to 97.2% w/w.

UPSTREAM
Lithium ripped nearly 10% in a week, and forecasters are already calling a Q3/Q4 peak

China's domestic lithium-carbonate futures went from a ~136,000 RMB/t low in early August to a 149,700 RMB/t close on Aug 12, briefly breaking 150,000 RMB/t intraday — a roughly 10% weekly move that stands in sharp contrast to the "cooling, choppy" tape described in this newsletter as recently as last week (SunSirs). The domestic spot average moved in lockstep, up 3,250 RMB/t in a single day to 148,000 RMB/t.

Multiple institutions are now calling for 150,000–200,000 RMB/t through the second half of 2026, with the year's price peak expected between the end of Q3 and the start of Q4 — a rebound thesis that had been a minority view just weeks ago when prices were still sliding. This is the kind of move that eventually shows up in cell and pack pricing; the BIS black-mass export rule (Policy) adds a second, policy-driven supply constraint on top of a market that just re-priced itself higher.

• Chinese lithium-carbonate futures: 136,000 → 149,700 RMB/t (Aug 12 close), ~10% weekly gain.

• Spot average: 148,000 RMB/t (Aug 12, +3,250 RMB/t d/d).

• Forecasters now call for 150,000–200,000 RMB/t in H2 2026, peak expected Q3-end/Q4-start.

• [NEED: source] citable USD/kg primary print from Benchmark/Fastmarkets (paywalled, no confirmable in-window figure).

MIDSTREAM
Sodium-ion's cost-parity clock and a defense-driven gigafactory groundbreaking

CATL reiterated this week that its sodium-ion cells will hit cost parity with LFP by end-2026, with full storage-system parity (cells plus integration) following in 2027 — a date that's held steady across multiple public statements rather than slipping, which is itself notable given how often chemistry-cost timelines drift (Renewables Now interview). The TENER Sodium BESS launched in Munich in June is now rated for 15,000 cycles at 25°C while retaining 70% capacity — a 25-30 year functional life that matters directly for BMS and warranty design, the layer where this newsletter's systems lens adds the most.

Separately, Forge Nano broke ground Aug 19 on "America's Battery Gigafactory" in Morrisville, North Carolina, explicitly framed around reducing U.S. dependence on foreign-controlled battery supply chains for defense applications — a domestic-manufacturing signal that lands the same week as the lithium price spike and the BIS black-mass rule, all pointing the same direction (TMCnet/GlobeNewswire). Capacity and funding figures for the Forge Nano project weren't confirmable this session; flagged for manual fill.

• CATL: sodium-ion/LFP cell cost parity by end-2026; full system parity in 2027; ~1 GWh cumulative shipments targeted by end-2026.

• CATL TENER Sodium BESS: 15,000-cycle rated life at 25°C, 70% capacity retention (~25-30 year service life).

• Forge Nano broke ground Aug 19 on a defense-focused US battery gigafactory (Morrisville, NC); [NEED: source] for capacity/funding figures.

DOWNSTREAM
Hyperscalers can now buy residential battery capacity directly, and the tax credit that used to subsidize it is gone

Sunrun and Voltus announced a multi-year "Bring Your Own Capacity" partnership on Aug 18 to deploy residential solar-plus-battery systems across PJM and MISO as distributed capacity explicitly aimed at AI data-center demand — the same causal chain (compute buildout → power demand → storage response) tracked in this newsletter's Macro backbone. Google already committed to a 3-year, up-to-100 MW BYOC deal with Voltus in June, and a New York compensation example cited $586,000/MW-year in potential revenue and savings for participating households (ESS News).

That commercial pull is arriving without federal help: the Section 25D residential credit (30% of system cost) is gone for any system placed in service in 2026 or later — a $3,600–4,500 swing on a typical $12–15K install (Rewiring America). Yet the segment's underlying demand looks durable: Sunrun's Q2 2026 storage attachment rate hit a company-record 74% (up from 70% a year earlier, reported Aug 5), with the installed base now at 266,000+ systems and ~4.6 GWh — standing context for a spotlight lane that just gained a second revenue channel (hyperscaler-funded VPP capacity) precisely as the first one (the tax credit) disappeared.

• Sunrun/Voltus "Bring Your Own Capacity" deal (Aug 18): residential batteries sold as distributed capacity to AI hyperscalers across PJM/MISO; Google's parallel Voltus deal covers up to 100 MW.

• Section 25D credit terminated for 2026+ installs — a $3,600–4,500 swing on a typical system, with no federal replacement.

• Standing context: Sunrun's Q2 2026 storage attach rate hit a record 74% (reported Aug 5), 4.6 GWh installed across 266,000+ systems.

POLICY
The compliance bar for domestic content just got tighter, not looser

This week's clearest policy read is on FEOC: the non-prohibited-foreign-entity content threshold for 45X/45Y/48E credit eligibility is now confirmed at a minimum 55% in 2026 (not the 40% figure referenced in earlier implementation guidance), stepping up 5 points a year to 75% by 2029 — a stricter bar than developers had been modeling, layered onto the 25% Section 301 tariff on lithium-ion battery cells and natural graphite already in effect since January (Morgan Lewis/IndexBox). The reported cost impact is severe: pack costs for FEOC-linked sourcing can rise as much as 69%, on top of forfeiting the ITC entirely if the content threshold isn't met.

The BIS black-mass/tungsten export rule enters its final week before the Aug 27 effective date, with no new developments this week beyond the countdown itself — but the compounding effect across all three levers (FEOC content, Section 301 tariffs, the black-mass export lock) is the same: every actor in the value chain, from residential installers to grid-scale BESS developers to recyclers, is now underwriting a materially higher domestic-content and compliance cost than they were pricing in even a quarter ago.

• FEOC non-PFE content threshold: confirmed at ≥55% for 2026 (up from earlier 40% guidance), stepping to 75% by 2029.

• Section 301 tariffs: 25% on lithium-ion non-EV cells and natural graphite, effective since Jan 1, 2026.

• Reported pack-cost impact from FEOC-linked sourcing: up to +69%, plus full ITC forfeiture risk.

CLOSING
If hyperscalers are now bidding for residential battery capacity directly, who else gets paid?

This week's clearest tension: Sunrun and Voltus just opened a channel for AI data-center operators to fund residential VPP capacity directly (mirroring Google's parallel 100 MW Voltus deal), arriving exactly as the federal tax credit that used to subsidize those same batteries disappeared — while FEOC's tightened 55% domestic-content threshold and a 10% lithium price spike both push the cost of building that hardware higher. Something has to give between "hyperscalers will now pay homeowners directly for grid capacity" and "batteries just got structurally more expensive to build domestically."

• Sunrun/Voltus BYOC deal (Aug 18): residential batteries now sellable as hyperscaler-funded VPP capacity in PJM/MISO.

• FEOC's 2026 non-PFE threshold is confirmed at 55%, tighter than earlier guidance — stacked on a 25% battery-cell tariff and a 10% weekly lithium spike.

• Reply and tell me: does hyperscaler-funded VPP capacity replace the lost 25D tax credit as residential storage's economic engine, or is it too small (100 MW-scale deals) to matter next to a mass-market subsidy that's gone?

Until next week,
BatteryJoy