UPDATED GROWTH PORTFOLIO EXPOSURE Performance: • Growth Portfolio: 431% (93% CAGR) • $QQQ: 171% • $ARKK: 154% • $SPY: 91% Exposure: • AI Hardware | 22% • Space Economy | 20% • Digital Platforms | 13% • Healthcare | 9% • Physical AI | 9% • AI Power | 7% • Drones | 7% • Quantum | 6% • Power Grid | 1%

$ONDS formed Ondas Sentinel as a new U.S. defense division that combines World View and DZYNE. The goal is to create a scalable defense platform that can pursue larger programs and deliver more integrated autonomous defense solutions.

$MU and $F signed a long-term memory and storage supply agreement for Ford’s next-generation vehicles. The deal gives Ford more secure automotive memory supply while Micron expands advanced DRAM production in the U.S. including at its Virginia fab.

$WULF signed a 20-year, nearly $20B deal with Anthropic for its Justified Data campus in Kentucky. The campus will support ~401MW of critical IT load with initial capacity online in H2 2027 and full capacity expected by early 2028.

$AVGO and $AAPL are extending their long-standing chip partnership through 2031. Broadcom will develop and supply custom ASIC silicon for multiple generations of Apple products under new multi-year agreements.

Morgan Stanley says the AI trade is rotating away from the stocks that led the rally with indexes likely to stay pressured as the “momentum unwind” hits larger index names. The firm now favors $MSFT, $AMZN and $META over chipmakers arguing hyperscalers remain attractive AI plays because of their strong core businesses.

B. Riley says Amazon’s RNG and OpenAI’s MRC protocols create a “major structural risk” for optical infrastructure and photonics stocks by flattening AI data center networks. The concern is that these architectures reduce the need for multi-tier switching and switch-to-switch optical links potentially cutting structural transceiver demand by 40% to 50%. The firm says $AAOI is “especially vulnerable” because $AMZN and $ORCL were expected to be key customers behind the 800G and 1.6T ramp that could push AAOI quarterly revenue toward ~$1B by 2H27.

$ONDS is acquiring DZYNE Technologies in an $876M adding autonomous aircraft, drone & counter drone capabilities to its defense autonomy platform. DZYNE is expected to generate $191M in 2026 revenue lifting Ondas 2026 revenue target to $525M after its DZYNE and Ominisys deals.

10 WAYS TO PLAY ROBOTICS IN 2026 1. $TER tests AI chips & owns the cobots through Universal Robots 2. $NOVT precision photonics & motion inside surgical & industrial robots 3. $OUST eyes & perception for physical AI machines that need to see & map physical world 4. $HIMX provides vision & display chips that help robots process images, depth & sensor inputs at edge 5. $AMBQ low power edge AI layer for robots & devices that need intelligence without burning massive amounts of energy 6. $ON becomes physical AI edge stack after acquiring $SYNA combining power, sensing, connectivity & edge AI for robotics 7. $SYM warehouse automation pure play using robotics & software to redesign how goods move through fulfillment centers 8. $VPG sits in force sensing & precision measurement layer that helps robots understand pressure, weight & real world movement 9. $TSLA humanoid platform bet with Optimus combining custom hardware, AI inference, autonomy software & real world manufacturing scale 10. $CCXI gives public market exposure to Agility Robotics which is one of clearest pure play humanoid robotics companies through its Digit warehouse robot

$MU is up more than 7% in overnight trading.

U.S. stock futures:

$MU broke ground on a $9.3B AI memory expansion in Hiroshima backed by up to $3.1B in Japanese government support. The facility will produce next-generation HBM for AI chips starting around summer 2028 expanding Micron’s global AI memory footprint.

Leaked Anthropic docs reportedly point to a 1.4GW Australia capacity push worth ~$22B. $IREN is one of the few names with announced gigawatt-scale Australian ambitions, 800MW Bundey campus with its connection agreement already secured and a state government partnership.

This is Peter Lynch’s favorite metric and I like it too because it ties valuation to growth. I prefer using a 2026 to 2028 window so the multiple adjusts for how fast the business is actually compounding. For example, $SOFI may trade at ~26x forward earnings but earnings are expected to grow from $0.60 in 2026 to $1.05 in 2028 implying a ~30% earnings CAGR and a PEG of ~0.9x.

POPULAR STOCKS BY PEG RATIO PEG < 1 usually means mispriced growth PEG > 2 starts to push into the danger zone Here’s how they stack up: • $TSLA 4.5x • $SPCX 3.8x • $GOOGL 2.9x • $AAPL 2.6x • $INTC 2.5x • $HOOD 2.0x • $PLTR 2.0x • $META 1.5x • $ASML 1.5x • $AMZN 1.4x • $MSFT 1.2x • $TSM 1.1x • $SOFI 0.9x • $HIMS 0.9x • $NOW 0.7x • $NVDA 0.6x • $AMD 0.6x • $AVGO 0.5x • $ORCL 0.4x • $MU 0.1x • $SNDK 0.1x

$SOFI took 11 years to reach its first 4M members and added nearly another 4M in the past year alone. Thats why the top 10 bank thesis still feels early since SoFi holds less than 15% of the deposit base of today's number 10 U.S. bank while building a one-stop financial platform with 30% member growth. The kicker is LPB (Loan Platform Business) which is the asset-light layer where SoFi earns a ~4.6% take rate on third-party originations with zero balance-sheet usage and zero credit risk. That creates a high margin, fee based lending revenue stream running at ~$12B annualized origination pace while SoFi still trades around 15x 2028 earnings.

$AMD could approach $100B in annual revenue by 2028 as $META, $ORCL and OpenAI commitments turn MI450 and Helios into a real second source for AI compute. The GPU deals are what re-rated AMD but the CPU bottleneck gives the story a second leg because every AI cluster still needs general compute for orchestration and inference workloads. That's why the CPU leg keeps compounding since agentic AI is making hyperscalers more CPU hungry with EPYC already at 46% server CPU revenue share and AMD's next-gen Venice already ramping on $TSM 2nm.

$GOOGL profit is expected to nearly double by 2028. When compute is the scarcest input in the economy then the company with the lowest internal cost per token wins out. Google is the only hyperscaler running its frontier model end to end on its own silicon with Gemini trained entirely on TPUs.

SK Hynix is expected to hit a 91% DRAM gross margin in Q2 showing how extreme this memory cycle has become. Now the company is set to list on Nasdaq this Friday under $SKHY giving U.S. investors a second direct way to play the AI memory boom besides $MU.