$MXL keeps adding to the story. Trinity platform just dropped today. URX850 SoC delivering carrier-grade bidirectional wireless backhaul at up to 10Gbps. What used to require multiple separate components, switching, QoS, wireless link aggregation, encryption, and timing functions, now sits on a single chip. Bharti Airtel, one of India’s largest carriers, is already on record saying they’ve been waiting for exactly this kind of integration. The part worth understanding is that this has nothing to do with the optical interconnect story I was posting about earlier. That’s data centers. This is cell towers connecting to core networks. Two completely separate markets, both accelerating, both running on $MXL silicon. Keystone PAM4 DSP for 400G and 800G optical transceivers in AI data centers. Rushmore targeting 1.6T. Panther going after memory latency. And now Trinity for 5G wireless backhaul at 10Gbps. The product roadmap keeps expanding into infrastructure layers that aren’t slowing down. OEM products based on Trinity expected H1 2027. URX850 already deployed globally across major carrier networks. Available now. This is what a picks and shovels position in multiple infrastructure cycles looks like. 🔗maxlinear.com/news/press-rel…

I started buying $LWLG about a year ago, and it’s already up around 10x. The risk is still there though. That hasn’t changed. It’s still early, not proven at scale, and revenue hasn’t caught up yet. But what has changed is the progress. For a long time it felt like one of those “cool tech, but when does it actually happen” stories. Lately that’s starting to shift. They’re not just talking about the tech anymore. It’s starting to show up inside real foundry flows like GlobalFoundries and Tower. Once something gets into a PDK and shows up in tape outs, it’s a totally different ballgame. People can actually start designing around it. On top of that, some customers are moving past early testing and getting closer to real products. That’s what I’m really watching. Not the headlines. Just whether this tech actually ends up inside real systems. So what exactly is $LWLG’s technology? Thread🧵👇

Yes, this is $OPTT. And what’s happening here is a lot more interesting than just a Coast Guard surveillance story. @OceanPowerTech Ocean Power Technologies just deployed its first PowerBuoy under a $6.5M DHS contract off the California coast last week. Each buoy sits offshore indefinitely, self powered by wave energy, feeding sensor data into Anduril’s Lattice command and control system alongside surveillance towers already in the water. No shore cables. No fuel. No crew. Just persistent eyes on the ocean running continuously. The part most people miss is the architecture. $OPTT isn’t just selling hardware to the government. It’s becoming a node in a defense intelligence network that #Anduril is building. Every buoy feeds into Lattice, and the more buoys go in the water, the more valuable the whole network gets. You don’t really replace that once it’s built in. And the use case isn’t going anywhere either. Migrant smuggling, drug trafficking, adversarial maritime activity off the U.S. coastline. The Coast Guard can’t put manned assets everywhere. Wave-powered buoys that run indefinitely and feed straight into AI driven command systems are a pretty obvious answer to that problem, and nobody else is really doing this at scale. Backlog up 165% year over year to ~$19.9M. Pipeline sitting at ~$163.9M. First buoy in the water last week. Financials are still a mess, but the #Anduril integration is what makes this story different from a typical small defense contractor.

$ABCL Called the accumulation phase. The stock is coming back. When I posted about treating the post-Phase 1 sell off as an opportunity, the thesis was simple. The data wasn’t bad. The market just wanted efficacy numbers that Phase 1 isn’t designed to deliver. Everything that actually mattered came back clean. And since then the story has only gotten stronger. Phase 1 data confirmed zero liver toxicity across all doses from 30mg to 900mg. Zero serious adverse events. Strong NK3R target engagement confirmed. Half life of approximately 24 days supporting once monthly dosing. Phase 2 already enrolling. Q3 topline data on track. Then on May 18, an insider made a fresh share purchase. Goldman sitting on 4.3M shares. Baker Bros increased from 9.1% to 10.8% the day before the readout. Cantor Fitzgerald initiated Overweight. Evan Seigerman reiterated Buy with $7 price target citing strong Phase 1 data and platform strength. Analyst consensus sitting around $9.00 against a current price that’s been recovering. Revenue doubled year over year in Q1. Net loss narrowed. $655M in liquidity. Four programs in clinical or IND enabling stage. ABCL575 Phase 1 data coming H2 2026. The pipeline isn’t #ABCL635 alone. Q3 is still the moment. The setup going into it keeps getting cleaner. 😉

Everyone’s chasing semiconductors, space, and AI right now, and honestly same, I’ve been deep in all of it. But there’s one company I keep coming back to quietly, one that’s building something in a completely different direction. That’s $DMRC Digimarc, and let me explain why it’s been on my radar. They do digital watermarking. Embedding invisible identification into content so you can prove what it is, where it came from, and whether it’s been tampered with. They’ve been doing this for nearly 30 years, and one of their longest running deployments is with a consortium of the world’s central banks to deter currency counterfeiting. Not a new idea for them at all. And honestly the reason this feels more urgent now is pretty obvious when you think about it. AI generated content is everywhere, deepfakes, synthetic images, AI written text, and the question of is this real is getting harder to answer every single day. Digital watermarking is one of the most viable solutions to that problem at scale, and $DMRC has been quietly building the infrastructure for exactly this for decades without much fanfare. The 2026 piece is worth paying attention to though. The EU AI Act comes into full force in May 2026, and it mandates watermarking and labeling of AI generated content across the board. Non compliance means fines up to 15 million EUR or 3% of global annual turnover, whichever is higher. This applies to any AI company operating in Europe, which is basically everyone. What kind of got me was this part. $DMRC developed the industry’s first digital watermarking solution compliant with C2PA 2.1, the standard the industry is converging around, and they co-chair the watermarking task force for that standard. So they’re not just selling into this space, they’re literally at the table where the rules are being written. That’s a different kind of positioning. The honest risk is the financials are still rough. Small cap, still burning cash, stock has been quiet for a while. The story is ahead of the numbers right now. But regulation moving in this direction with $DMRC sitting at the center of the standard, that’s not nothing. Staying on my watchlist for sure.

This looks small but I think it matters more than it seems for $RDW. The ~$12.8M is not the story. What stands out is this is both a first sale and a standard component selection. First sale means ELSA actually made it into a real program. For new space hardware that is usually the hardest step. But being baselined inside Moog’s METEOR platform is the bigger signal. It does not guarantee revenue, but it creates a path. Every time that platform is used ELSA now has a seat at the table. That is very different from a one off contract. Technically this lines up with where the market is going. Higher power density lower mass faster production. That is exactly what large satellite constellations and defense systems need. From an investment perspective this is not about a single contract. It is about positioning inside a platform, and that is where repeat revenue starts to show up. Feels like Redwire is starting to move from selling parts to getting embedded inside systems. Market might be missing that. $RDW

$ABCL dropped Phase 1 data and the stock sold off. I think that’s a mistake. ABCL635 hit every single checkpoint Phase 1 is supposed to hit. No liver toxicity at any dose. Zero serious adverse events. Testosterone suppression confirming strong NK3R target engagement. Half life of approximately 24 days, which is exactly what you need to support once monthly subcutaneous dosing. The pharmacokinetic profile came back clean. The tolerability profile came back clean. The company immediately advanced into Phase 2 on the strength of this data. The market sold it because Phase 1 doesn’t come with efficacy numbers. That’s a different thing from the data being bad. Phase 1 confirmation is what makes Phase 2 credible, and that’s exactly what this was. The piece that keeps coming back to me is the hepatotoxicity comparison. Veozah, the only approved oral NK3R drug, has a Black Box Warning for liver toxicity and still pulled $234M in 9 months. ABCL635 showed zero liver enzyme elevations across all doses tested from 30mg to 900mg in the Phase 1 data released last week. That’s the dosing headroom that Astellas never had. If higher doses are safe, you can push efficacy further than Veozah ever could. Goldman sitting on 4.3M shares. Baker Bros increased from 9.1% to 10.8% the day before the readout. These people saw the data before we did. They added going into the announcement. $655M in liquidity. EPS beat. Revenue beat. Phase 1 cleared. Phase 2 on track for Q3. The sell off gave us a better entry on a story that actually got cleaner this week, not worse. Treating this as an accumulation phase.

$INFQ has been running hard this week and the reason is pretty clear once you dig in. NVIDIA just launched Ising earlier this week. It is the world’s first open-source AI models built specifically for quantum computing. They’re designed for quantum processor calibration and error correction decoding, and they’re running 2.5x faster and 3x more accurate than traditional methods. Out of all the quantum companies NVIDIA highlighted, Infleqtion was the only one that showed up on both the calibration list and the decoding list. Citron Research jumped on it right away, calling it the most obvious mispricing in the entire quantum space right now. Their point was simple. Rigetti ($RGTI), which has roughly twice the market cap of $INFQ, didn’t make NVIDIA’s list at all. On top of that, Citi initiated coverage this week with a Buy rating and a $20 price target. The stock went from $9.81 at the end of March to an all time high of $21.28 on Friday. Basically an 80% run in under three weeks. Most pure play quantum companies are still betting entirely on a future that hasn’t arrived yet. $INFQ is structurally different, and I think most people still haven’t fully clocked it. More on that below 🧵👇

Every beam. Every day. Real progress. G2_Austin. America needs more energy infrastructure, and it’s being built right here.

$ABCL This is the context that matters going into today’s Phase 1 readout. Veozah, a Black Box Warning for hepatotoxicity and all, pulled $234M in 9 months in a market that’s still building. That’s not the ceiling, that’s what a drug with a Black Box Warning can do in a brand new market. The liver warning forced Astellas to cap the dose at 45mg even though Phase 2 data showed higher doses delivered better efficacy. They left effectiveness on the table because the compound itself was toxic, not the target. If #ABCL635 comes back with clean safety today, that changes the dosing calculus entirely. An antibody mechanism that hits the same NK3R target without the hepatotoxicity risk could theoretically be dosed higher and deliver meaningfully better efficacy than Veozah ever could. That’s a meaningfully better drug in the same market, not just a slightly cleaner version. Is most of the upside priced in right now? Maybe. Probably some of it. But I’ve been in this stock long enough to know that a down move on data, if it comes, would be an opportunity, not a reason to leave. The science here is real. The team is real. The platform behind ABCL635 has already proven it can move from discovery to clinic at a pace traditional pharma can’t match. Whatever happens today, I’m not going anywhere. This is a company I trust, and that’s honestly the only thing that matters when you’re holding through a binary event.😉

Called this one. Back when I posted about $MXL, it wasn’t getting a lot of attention outside of people already tracking the optical interconnect space. The Q1 numbers changed that conversation fast. Revenue up 43% year over year to -$137.2M. Infrastructure segment, which is the optical connectivity piece, up 136% year over year and now the largest end market. Stock was the best performing IT name in April according to Seeking Alpha, with the top 10 IT performers delivering gains ranging from 71% to 336%, momentum concentrated in semiconductors, photonics, and power electronics as AI infrastructure buildout accelerated. $MXL was sitting at the top of that list. Once you understand where MaxLinear actually sits in the stack, the move makes a lot of sense. They don’t make the transceiver. They make the semiconductor chips that go inside transceivers across multiple manufacturers. Whoever wins the transceiver market, $MXL supplies the silicon underneath. That’s not a bet on one customer or one design win. That’s a position in the layer that every optical interconnect solution depends on, and that layer is becoming more critical as AI clusters push bandwidth requirements to 800G and 1.6T. And the analyst moves post-earnings aren’t small either. Price targets getting moved to 105% above current levels isn’t a minor adjustment. That’s the market catching up to a thesis that was already in the numbers for anyone paying attention. Full year 2026 optical data center revenue guidance raised to $150M to -$170M. This isn’t a one quarter setup. Every major hyperscaler is still in the middle of a multi-year buildout and $MXL is in the supply chain for all of it. Still holding. Still watching. 🫣

Ondas reported record Q1 2026 results with $50.1 million in revenue, raised its 2026 revenue target to at least $390 million, and expanded pro forma backlog to $457 million. The quarter reflected accelerating demand across counter-UAS, ISR, and autonomous systems markets while advancing Ondas’ global operating platform and AI-enabled multi-domain ISR strategy. $ONDS ondas.com/post/ondas-reports…

$PENG Penguin Solutions is one of those companies that’s genuinely hard to explain in one sentence, and I think that’s part of why it’s still flying under the radar for a lot of people. Think of $PENG as the company that takes $NVDA ’s GPUs and actually builds the AI cluster around them. The design, the architecture, the deployment, the ongoing management. The hardware is someone else’s. The expertise of putting it all together is theirs. Here’s how I think about it. When a company, a government, a university, or a neocloud provider wants to build an AI cluster, they usually don’t have the in-house expertise to design, deploy, and operate it at scale. That’s where PENG comes in. 25 years of high performance computing experience, applied directly to AI factory buildouts. They handle the full stack, hardware selection, cluster architecture, deployment, and then ongoing managed services through their ICE ClusterWare platform. Over 3.3 billion hours of GPU runtime management under their belt. That’s not a number most companies can put on a slide. What I keep coming back to is the customer diversification story. The AI infrastructure market spent the last two years being dominated by hyperscaler spending, lumpy, unpredictable, and concentrated. #PENG is actively pivoting toward enterprise, neocloud, and sovereign AI customers. Governments across Europe and the Middle East are building localized AI infrastructure to keep data within their borders. They can’t just call AWS. They need a partner who can design and operate the whole system. That’s a stickier customer profile than hyperscalers, and it compounds differently. The software layer is the underappreciated piece. ICE ClusterWare is a proprietary cluster management platform purpose-built for AI workloads. That’s not commodity hardware reselling. It’s a software moat that generates managed service revenue on top of the hardware sale, and every dollar from managed services carries significantly higher margins than the hardware underneath it. Q2 in early April showed revenue down year over year but net income up sharply and full year guidance raised to 12% growth. Goldman Sachs initiated with a Buy this morning. Barclays was already there. 8 analysts, Strong Buy consensus, average target around $28. The honest risk is the transition period. Moving away from hyperscaler concentration toward enterprise and sovereign AI takes time, and the revenue line will be lumpy while that shift plays out. But 25 years of HPC expertise, a proprietary software platform, managed services that compound over time, and a pivot toward sovereign AI and enterprise at exactly the moment those markets are accelerating. That’s the setup I’m watching.

Tomorrow, $ABCL has its first Phase 1 readout for ABCL635. Most of the upside is priced in IMO. However, tomorrow's data will show: (i) if the mAb engages the target and it is not excluded by the blood-brain barrier; and (ii) preliminary safety which is very important (see below). The next milestone will be efficacy. I am doing a long article about ABCL635 for after the ph1 readout, but I wanted to share something. ABCL635 engages NK3r, same target as Fezolinetant (Veozah) from Astellas. The oral pill Veozah has made ~$234M in sales in the 9 months since launch in 2025, in a $4B VMS non-hormonal treatment market. REMEMBER that this molecule has a BLACK BOX WARNING for Hepatotoxicity!!! AND $234M in 9 months! The table below shows the results from Fezolinetant in ph 2b and ph3s. In phase 2 they tested QD (one pill a day) vs BID (two pills a day) each at high and low doses (30 and 120 mg for QD and 15 and 90 for BID). The endpoint was the difference in VMS frequency from baseline at 12 weeks. This means the reduction of episodes that the woman had of hot flashes compared to baseline (w/o the treatment). Example: -3 is better than -1, since -3 means they experienced 3 episodes less than baseline, while -1 means they experienced only 1 less. The trend is clear for ph2 and ph3. The molecule clearly works and engages the target (p-value < 0.05 in most cases) and higher doses resulted in fewer VMS episodes. GOOD!! However, they chose 45 mg of Fezolinetant for the Veozah pill, even if ph2 showed more reduction of VMS episodes at higher doses, WHY?? Higher dosage = Hepatotoxicity. 🚨My point: THERE IS A LOT OF EFFICIACY LEFT ON THE TABLE for ABCL635!!! The hepatotoxicity is most probably due to the Fezolinetant compound, not the NK3r target engaged. IF this is true and ABCL635 doesn't have toxicity and is safe, AbCellera could trial a safer higher dosing resulting in HIGHER EFFICACY!! If ABCL635 is safer and has more efficacy than Fezolinetant they would only compete over how convenient it is to have one pill a day or one shot a month, that is not up to me to decide. But still, it will be a clear contender for the $6B non-hormonal treatment VMS market projected for 2030 and could maaaaaaaaaybe prey on the Hormone replacement therapy market of $26.5B by 2030. Big IF, I know, but there is a good chance.