Sekėjai

Ieškoti šiame dienoraštyje

2026 m. rugpjūčio 19 d., trečiadienis

Baidu's Profit and Revenue Falter Amid Its Pivot to AI


“Baidu's financial performance continued to deteriorate in the second quarter as the Chinese internet titan works to remake itself into an artificial-intelligence company.

 

The Beijing company, once considered China's answer to Google, has been pouring money into AI, autonomous driving and chip design even as its bread-and-butter advertising business has continued to shrink.

 

 The investments have been slow to show results: Net profit more than halved for the third consecutive quarter, while revenue declined for more than a year.

 

Baidu reported a 68% drop in second-quarter net profit to 2.32 billion yuan, equivalent to $344.2 million. Revenue came to 31.32 billion yuan, down 4.2% from a year earlier.

 

The soft showing comes as Baidu's Hong Kong-listed shares have shed more than one-fifth of their value this year, underperforming the broader market as the company's legacy business continued to drag its outlook. Investors, meanwhile, are keeping a close eye on the listing progress of its chip unit, Kunlunxin, expected in Hong Kong within this year.

 

Baidu is pinning its hopes on its core AI-powered business, which the company said continued to account for half of general business revenue.

 

Fitch Ratings last week downgraded its long-term issuer default rating on Baidu to A-minus from A, citing the structural decline in the company's search ad business. It said it expects emerging AI search and competing chatbots to weigh on Baidu's search business monetization.

 

Baidu faces stiff competition in China from tech titans such as Alibaba and ByteDance, as well as startups such as DeepSeek and Moonshot AI in the AI sector.

 

For the three months ended June, Baidu's core AI-related business generated 25% revenue growth year over year, though the figure declined 8% from the prior quarter.

 

Our AI-powered business is now "firmly established as the core of Baidu," said Robin Li, the company's co-founder and chief executive. "We are strengthening the foundations for our next phase of AI-driven growth."” [1]

 

1. Baidu's Profit and Revenue Falter Amid Its Pivot to AI. Qu, Tracy.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B4.  

„Unitree“ IPO – pirmasis Kinijos robotikos sektoriuje – trečiadienį humanoidų ir robotų bendrovė Šanchajuje pradės savo veiklą viešai


„Būsimas humanoidų ir robotų startuolio „Unitree“ kotiravimas gali sukelti lėšų rinkimo ažiotažą visame Kinijos robotikos sektoriuje, o debiutas bus pagrindinis rinkos apetito išbandymas.

 

Hangdžou įsikūrusi „Unitree“, oficialiai žinoma kaip „Yushu Technology“, trečiadienį taps pirmąja humanoidų ir robotų bendrove, prekiaujančia žemyninėje Kinijos dalyje, kai jos akcijos bus kotiruojamos Šanchajaus „Nasdaq“ tipo STAR rinkoje.

 

„Unitree“ IPO yra reikšmingas, nes jis suteikia svarbų A klasės akcijų vertinimo etaloną įkūnytajam dirbtiniam intelektui ir humanoidų robotikai“, – sakė „Morningstar“ analitikas Kangyuxiao Li.

 

Pasiūlymas sulaukė didžiulės paklausos ir galėtų atverti kelią naujiems IPO.

 

Mažmeniniai investuotojai suskubo užsitikrinti akcijas, pateikdami 9,8 mln. užsakymų – daugiau nei praėjusį mėnesį sulaukė atminties lustų gamintojos CXMT didelio pasisekimo sulaukęs pasiūlymas.

 

„Unitree“ teigė, kad mažmeninė jos pasiūlymo dalis sudarė daugiau nei 5 500. prenumeratos kartų.

 

Kitos Kinijos robotikos įmonės prekiauja Honkonge, tačiau „Unitree“ įtraukimas į biržą suteikia žemyninės Kinijos investuotojams tiesioginę prieigą prie vienos iš pirmaujančių sektoriaus įmonių, sakė Li, pridurdamas, kad tai gali turėti įtakos būsimų robotikos IPO ir privačių įmonių vertinimams.

 

Apie 20 % siūlomų „Unitree“ akcijų buvo paskirstyta strateginiams investuotojams, įskaitant dirbtinio intelekto laboratoriją „DeepSeek“ ir didelių valstybinių įmonių, tokių kaip „China Southern Power Grid“, investicinius padalinius.

 

Pekinui nukreipiant išteklius į tokius sektorius kaip robotika, „Unitree“ tapo vietine humanoidinių robotų čempione.

 

Technologijų sunkiasvorių, tokių kaip „Meituan“ ir rizikos kapitalo įmonės HSG, remiamas startuolis patraukė nacionalinį dėmesį, kai jo humanoidiniai robotai pernai per Mėnulio Naujųjų metų šventę atliko liaudies šokį.

 

„Unitree“ pritraukė 900 mln. JAV dolerių, įkainojusi akcijas po 150,80 juanių už vienetą, o tai reiškia, kad jų vertė siekia apie 9,1 mlrd. JAV dolerių.

 

Įkurta 2016 m., bendrovė specializuojasi kuriant robotus, kurie vaikšto ir bėga kaip žmonės. Pirmadienį ji paskelbė apie savo naujausią kūrinį: „Superžmogus robotas“, galintis iš vietos pašokti 2 metrus (daugiau nei šešias pėdas) ir pasiekti maksimalų 12,66 metro per sekundę greitį – abu šie rodikliai viršija žmonių pasaulio rekordus.

 

„Unitree“ nuolat augino pajamas ir pelną, tačiau dėl augančių mokslinių tyrimų ir plėtros išlaidų jos pelnui spaudimą turėjo augančios išlaidos.

 

Pirmąjį šių metų ketvirtį pajamos išaugo 68 %, o pakoreguotas grynasis pelnas sumažėjo daugiau nei 50 % dėl mokslinių tyrimų ir plėtros išlaidų.

 

„Unitree“ pelnas artimiausiu metu, greičiausiai, dar labiau sumažės, nes ji daugiau investuoja į robotų „smegenų“ kūrimą – sritį, į kurią dar neinvestavo daug pinigų, teigė „Counterpoint Research“ atstovas Wei Sunas. [1]

 

Ar yra kokių nors „Unitree“ robotų, galinčių atlikti namų ruošos darbus? Kiek tai kainuoja?

 

Taip, „Unitree“ robotai plačiai reklamuojami, kaip galintys atlikti namų ruošos darbus, tačiau jie dar nėra visiškai paruošti naudoti namų pagalbininkai. Pagrindinis modelis, galintis atlikti šias užduotis, yra „Unitree G1 Humanoid Robot“, kurio pradinė kaina tiesiogiai iš gamintojo prasideda nuo 13 500 USD.

 

Nors „Unitree“ dažnai leidžia reklaminius vaizdo įrašus, kuriuose rodoma, kaip „G1“ atlieka įspūdingas užduotis, tokias, kaip riešutų skaldymas, skrebučių kepimas, butelių atidarymas ir skalbinių lankstymas, įsigijus pagrindinę mažmeninę versiją, virtuvės stalviršiai nebus automatiškai išvalyti.

Štai kaip suskirstomos kainos, modeliai ir realios galimybės:

„Unitree G1“ konfigūracijos

• „Unitree G1 Basic“ (13 500–16 000 USD): tai bazinė vartotojams skirta versija. Ji turi 23 laisvės laipsnius (DoF) ir apima pagrindines dirbtinio intelekto funkcijas, tačiau nepalaiko antrinio programavimo ar programinės įrangos kūrimo. Ji dažniausiai tarnauja, kaip iš anksto užprogramuota aparatinės įrangos pusiausvyros ir pagrindinio judrumo demonstracija.

 

• „Unitree G1 EDU“ (43 500–73 900 USD): jei iš tikrųjų norite programuoti pasirinktinį elgesį (pvz., siurbimą ar gėrimų patiekimą), turite pereiti prie, kūrėjams skirto, EDU modelio. Šioje versijoje sąnarių judesių amplitudė padidėja iki 43 laisvės laipsnių, įdiegtos itin miklios robotinės rankos ir suteikiama prieiga prie SDK.

 

Alternatyvus biudžeto modelis

• „Unitree R1“ (4 900–5 900 USD): „Unitree“ taip pat siūlo itin prieinamą humanoidą, vadinamą R1. Nors tai yra nebrangios inžinerijos stebuklas, baziniame modelyje visiškai nėra rankų / griebtuvų ir jis negali atlikti fizinių namų manipuliacijų.

 

Realybės patikrinimas

Didžioji dauguma „Unitree“ humanoidų pardavimų (apie 74 %) atitenka tik universitetams ir dirbtinio intelekto kūrėjams. Dabartinėje vartotojų robotikos būklėje kiekvienas darbas, kurį matote atliekant humanoidą reklaminiame vaizdo įraše, yra kruopščiai apmokytas sustiprinto mokymosi būdu arba rankiniu būdu parašytas scenarijaus forma. Robotams vis dar trūksta apibendrinto „smegenų“ efektyvumo, kad jie patys suprastų neaiškią komandą, pvz., „eik sutvarkyti netvarkingą svetainę“. Be to, aktyvus akumuliatoriaus veikimo laikas paprastai yra ribotas iki maždaug 1–2 valandų, kol jį reikia pakeisti ar įkrauti.

 

Ar dirbtinis intelektas, naudojamas „Unitree G1 EDU“, yra su atvirais svoriais?

 

Taip, „Unitree“ teikia atvirojo kodo įkūnytus dirbtinio intelekto modelius, duomenų rinkinius ir kūrimo paslaugas „Unitree G1“ ir EDU platformų valdymo sistemoms.

 

Atvirojo kodo dirbtinio intelekto pasiūlymai

• Modeliai ir sistemos: „Unitree“ dalijasi pagrindiniais imitacijos mokymosi modeliais (pvz., difuzijos politika ir ACT) ir tokiomis sistemomis, kaip „unitree_il_lerobot“, skirtomis antriniam kūrimui.

 

• Duomenų rinkiniai: viešai prieinami duomenys apima manipuliavimo duomenų rinkinius, skirtus G1 griebtuvui ir rankų vikrumui.

 

• Judėjimas ir valdymas: tyrimų paketai ir judesio sekimo duomenų rinkiniai leidžia kūrėjams modifikuoti ir perkvalifikuoti valdymo būdus.

 

1. Unitree IPO to Mark a First For China's Robotics Sector --- Humanoid-robotics company is set to go public in Shanghai on Wednesday. Qu, Tracy.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B4. 

Unitree IPO to Mark a First For China's Robotics Sector --- Humanoid-robotics company is set to go public in Shanghai on Wednesday


“A coming listing by humanoid-robot startup Unitree could unleash a fundraising frenzy across China's robotics sector, with the debut serving as a key test of market appetite.

 

Hangzhou-based Unitree, officially known as Yushu Technology, is set to become the first humanoid-robotics company to trade on the mainland when it goes public on Shanghai's Nasdaq-like STAR Market on Wednesday.

 

"Unitree's IPO is significant because it provides an important A-share valuation benchmark for embodied AI and humanoid robotics," Morningstar analyst Kangyuxiao Li said.

 

The offering has drawn huge demand and could pave the way for more IPOs.

 

Retail investors have rushed to secure shares, submitting 9.8 million orders -- more than memory chip maker CXMT's blockbuster offering drew last month.

 

Unitree said the retail tranche of its offering was more than 5,500 times subscribed.

 

Other Chinese robotics firms trade in Hong Kong, but Unitree's listing gives mainland investors direct exposure to one of the sector's leading companies, Li said, adding that it could influence the valuations of future robotics IPOs and private companies.

 

About 20% of Unitree shares on offer have been allocated to strategic investors including AI lab DeepSeek, and the investment arms of major state-owned enterprises such as China Southern Power Grid.

 

As Beijing channels resources into sectors like robotics, Unitree has emerged as a homegrown champion in humanoid robots.

 

Backed by tech heavyweights like Meituan and venture-capital firm HSG, the startup captured national attention when its humanoid robots performed a folk dance during a Lunar New Year Gala broadcast last year.

 

Unitree has raised $900 million after pricing shares at 150.80 yuan apiece, implying a valuation of about $9.1 billion.

 

Founded in 2016, the company specializes in developing robots that walk and run like humans. On Monday, it announced its latest creation: the "SuperMan Robot", which can jump 2 meters (more than six feet) from a standstill and hit a peak running speed of 12.66 meters per second, both surpassing human world records.

 

Unitree has been steadily growing revenue and profits over the years, but its bottom line has come under pressure from ballooning research-and-development expenses.

 

Revenue surged 68% in the first quarter of this year, while adjusted net profit slumped more than 50% because of R&D costs.

 

Unitree's profit is likely to drop further in the near term as it invests more to develop "brains" for robots, an area it hasn't poured much money into yet, Counterpoint Research's Wei Sun said.” [1]

 

Any Unitree's robots that can do household chores? How much does it cost?

 

Yes, Unitree’s robots are heavily marketed as being able to do household chores, but they are not yet out-of-the-box turnkey home assistants. The primary model capable of these tasks is the Unitree G1 Humanoid Robot, which has a starting price of $13,500 directly from the manufacturer.

While Unitree frequently releases promotional videos showing the G1 executing impressive tasks like cracking nuts, making toast, opening bottles, and folding laundry, buying the basic retail version will not automatically clear your kitchen counters.

Here is how the pricing, models, and real-world capabilities break down:

Unitree G1 Configurations

           Unitree G1 Basic ($13,500 - $16,000): This is the base consumer version. It features 23 degrees of freedom (DoF) and includes basic AI functionality, but it does not support secondary programming or software development. It mostly serves as a pre-programmed showcase of the hardware's balance and basic agility.

           Unitree G1 EDU ($43,500 - $73,900): If you actually want to program custom behaviors (like vacuuming or serving drinks), you have to step up to the developer-focused EDU model. This version increases the joints up to 43 DoF, includes highly dexterous robotic hands, and grants access to the SDK.

Alternative Budget Model

           Unitree R1 ($4,900 - $5,900): Unitree also offers an ultra-affordable humanoid called the R1. While it is a marvel of low-cost engineering, the base model completely lacks hands/grippers and cannot perform physical household manipulation.

The Reality Check

The vast majority of Unitree’s humanoid sales (around 74%) go strictly to universities and AI developers. In the current state of consumer robotics, every chore you see a humanoid perform in a promo video has been carefully trained via reinforcement learning or manually scripted. The robots still lack the generalized "brain" efficiency to understand a vague command like "go clean up the messy living room" on their own. Furthermore, active battery life is typically limited to about 1 to 2 hours before needing a swap or a charge.

 

Is the AI used in Unitree G1 EDU open weights?

 

Yes, Unitree provides open-source embodied AI models, datasets, and development frameworks for the Unitree G1 and EDU platforms.

Open-Source AI Offerings

           Models & Frameworks: Unitree shares mainstream imitation learning models (like Diffusion Policy and ACT) and frameworks such as unitree_il_lerobot for secondary development.

           Datasets: Publicly available data includes manipulation datasets for the G1 gripper and dexterity hands.

           Locomotion & Control: Research stacks and motion tracking datasets allow developers to modify and retrain control policies.

 

1. Unitree IPO to Mark a First For China's Robotics Sector --- Humanoid-robotics company is set to go public in Shanghai on Wednesday. Qu, Tracy.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B4. 

Naujos darbo vietos seniems dirbtinio intelekto lustams


„Kiek vertas senas dirbtinio intelekto lustas? Atsakymas į šį klausimą turi didelę reikšmę. Debesų kompiuterijos žaidėja „CoreWeave“ neseniai užsiminė, kad tai gali būti daug daugiau, nei daugelis žmonių manė.

 

Bendrovė neseniai pasirašė sutartį dėl „Nvidia A100“ lustų nuomos „patrauklia kaina“ iki 2029 m., praėjusią savaitę kalbėdamas su analitikais sakė „CoreWeave“ finansų vadovas Nitinas Agrawalas. Šie lustai buvo išleisti 2020 m. ir buvo gaminami apie trejus metus. „Vis dažniau matome ilgesnį jų naudojimą už didesnes kainas“, – sakė jis, – „o tai suteikia potencialą dideliam tolesniam augimui“.

 

Tai gali būti svarbu, nes įmonės paprastai savo lustus amortizuoja per ketverius–šešerius metus. Šis vertės sumažėjimas turi atsispindėti jų pelno (nuostolių) ataskaitose, paveikdamas pelną. Tačiau pelno mažinimo poveikis išnyksta, kai lustai visiškai amortizuojami. Tai reiškia, kad pardavus prieigą prie senų lustų, galima gauti didelę pelno maržą – jei tik kaina, kurią įmonės gali nustatyti, per daug nesumažėja.

 

A Tai, kad devynerių metų senumo lustas vis dar veikia ir generuoja pardavimus, suteikia tam tikro pagrindo manyti, kad lustai yra perspektyvus turtas, už kurį galima skolinti. Bankai ir turto valdytojai gali būti pasirengę teikti kreditus „CoreWeave“ ir kitoms debesijos kompiuterijos įmonėms mažesnėmis palūkanų normomis, įsitikinę, kad jų lustų užstatas kurį laiką netaps bevertis.

 

Žinoma, yra ir priešingų argumentų. Pirmasis yra tas, kad daugelis dirbtinio intelekto lustų sugenda. Duomenų apie tai, kaip dažnai jie sugenda, trūksta, tačiau „Meta Platforms“ 2024 m. ataskaitoje apie 54 dienų trukmės „Llama 3“ modelio mokymą teigiama, kad gedimų rodiklis gali būti apie 9 % per metus. Darant prielaidą, kad šis rodiklis laikui bėgant didėja, devynerių metų senumo dirbtinio intelekto skaičiavimo klasteris dėl gedimų gali lengvai būti gerokai mažesnis nei pusė pradinio dydžio.

 

Seno lusto gyvybingumas vis labiau abejotinas, nes skubėjimas kurti dirbtinio intelekto duomenų centrus susiduria su energijos vartojimo apribojimais. Naujesnės kartos „Nvidia“ lustai yra daug efektyvesni energijos vartojimo požiūriu nei A100. Įmonės gali nuspręsti, kad svarbiau iš naujų lustų gauti daugiau dirbtinio intelekto skaičiavimo darbo, nei didinti pelną naudojant senus, mažiau efektyvius. „CoreWeave“ pareiškimas apie savo A100 palieka neatsakytą svarbų klausimą: ką tiksliai jie gali įkrauti? Nors Agrawal kainą apibūdino kaip patrauklią, neaišku, ką jis tuo turėjo omenyje. „Patrauklumas“ yra žiūrinčiojo akyse. Ir tai labiau nei kas kita nulems, kokią vertę išlaikys senesni lustai.“ [1]

 

1. New Jobs for Old AI Chips. Fitch, Asa.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B11.  

New Jobs for Old AI Chips


“How much is an old AI chip worth? The answer to that question has significant implications. Cloud-computing player CoreWeave recently suggested it could be a lot more than many people have been assuming.

 

The company recently signed a deal renting out Nvidia's A100 chips at "an attractive price" in a deal lasting into 2029, CoreWeave finance chief Nitin Agrawal said in a call with analysts last week. Those chips came out in 2020 and were produced for about three years. "Increasingly, we are seeing longer utilization at higher prices," he said, "offering the potential for significant further upside."

 

That could be a big deal because companies typically depreciate their chips over four to six years. That reduction in value has to run through their income statements, hitting profit. Yet the drag on the bottom line ceases after chips are fully depreciated. This means that selling access to old chips could come with high profit margins -- as long as the price companies can charge doesn't go down too much.

 

A nine-year-old chip still running and generating sales gives some credence to the idea that chips are viable assets to lend against. Banks and asset managers might be willing to extend credit to CoreWeave and other cloud-computing operations at cheaper interest rates, confident their chip collateral won't become worthless for a while.

 

Of course, there are counterpoints. The first is that many AI chips fail. Data on how frequently they fail is scarce, but a Meta Platforms report in 2024 on a 54-day training run for its Llama 3 model suggested the failure rate could be about 9% a year. Assuming that rate accelerates as time goes on, a nine-year-old AI computing cluster could easily be well below half its initial size because of failures.

 

The viability of an old chip is increasingly questionable as the rush to build out AI data centers runs into power constraints. Newer generations of Nvidia's chips are orders of magnitude more energy efficient than the A100. Companies may judge it more important to harvest more AI computational work from new chips than to boost profits by running old, less-efficient ones.

 

CoreWeave's statement about its A100s leaves an important question unanswered: What, exactly, is it able to charge? While Agrawal described the price as attractive, it is unclear what he meant by that. "Attractive" is in the eye of the beholder. And that, more than anything, will determine what value older chips retain.” [1]

 

1. New Jobs for Old AI Chips. Fitch, Asa.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B11.  

What's Next for Oil Companies?


“If there's any business for which changing course really is like turning around a supertanker, it's Big Oil.

 

At the beginning of this year, the Western world's five large, integrated energy companies were talking about things like "structural cost efficiencies," "consistent delivery" and "superior shareholder returns, despite declining oil prices." Those are hardly words that send investors' pulses racing.

 

Then, war with Iran delivered a price windfall more quickly than business plans could adjust to it. Compared with last year, the group is expected to generate about $75 billion more free cash flow in 2026. Figuring out what to do with extra money is a nice problem to have, but still a problem. Investors tend to think of dividends and buybacks as ways of returning cash to shareholders. Paying down borrowings is an equally valid route to enhancing owners' wealth.

 

BP stands out in addressing that part of its balance sheet. It was the most indebted of its peers at the beginning of this year, and new management pledged to fix that by selling assets and halting stock buybacks. During the second quarter, BP cut various financial liabilities, which include debt, hybrid debt, leases and settlement payments for the Deepwater Horizon disaster, by $7 billion.

 

Analysts surveyed by FactSet expect net debt for the five supermajors combined to fall by $70 billion between 2025 and 2028. The transformation at smaller pure-play drillers and refiners will be more dramatic. Expand Energy and Valero are seen having net cash on balance sheets by 2028.

 

Then what? It's tempting to drill, baby, drill, especially with exploration costs and shale-patch break-even prices becoming more reasonable. But sinking much more money than planned into major oil and gas projects is hard to justify unless there's a strong case that commodity prices can stay higher for longer. And drilling too much today could produce its own bust tomorrow.

 

In addition to buying back more of their own shares, companies could use cash to start buying one another. Deals tend to be announced when prices are moderate rather than historically high or low.

 

The last big supermajor takeover was Chevron's $53 billion deal for Hess, an all-stock deal announced in 2023 and completed last summer. A year earlier, Exxon closed its $60 billion acquisition of Pioneer Natural Resources, also for stock.

 

Healthy balance sheets make smaller all-cash acquisitions in the single-digit billions of dollars more likely. That may be especially tempting in the next 2 1/2 years, before the next president takes office. It's a window of opportunity for more consolidation while U.S. antitrust enforcement is still energy-industry friendly. One way or another, more cash is headed shareholders' way. Some will be gradual, but expect some to arrive in big chunks amid press releases touting "synergies" and "world class assets."” [1]

 

1. What's Next for Oil Companies? Jakab, Spencer.  Wall Street Journal, Eastern edition; New York, N.Y.. 19 Aug 2026: B11.

Europe Pushes for Moonshot Breakthroughs to Compete With China and U.S. Using SPRIND


"The shiny cylindrical tanks and slant-roofed control panels shoved against a wall in C3 Biotech’s laboratory have essentially been turned into a multimillion-dollar graveyard of high-grade stainless steel equipment.

 

All of this state-of-the art gear was discarded during the company’s quest to discover a cheap way to convert agricultural waste like potato peels or sugar beet pulp into acrylic glass, a material widely used in manufacturing.

 

Nigel Scrutton, a co-founder of C3 Biotech in Stockport, England, makes a point of showing off these expensive failures before explaining that the eureka solution involved using discount plastic tubs — the kind you might find at Walmart.

 

What is equally surprising is that this British company’s breakthrough has been funded by an unusual grant from the German government, part of a novel effort to support shoot-the-moon innovation throughout Europe.

 

The pressure on Europe to better compete with Chinese and American technological and industrial advancements is peaking.

 

Europe has been trailing the world’s two largest economies for more than a decade on several measures of competitiveness, including investment, research and development, and productivity growth.

 

At stake is Europe’s “lead in many historic strongholds, including automotive, pharmaceuticals and machinery,” a report on competitiveness from McKinsey Global Institute concluded.

 

The region would have to spend nearly $1 trillion more in investment annually just to keep up, according to estimates.

 

Soaring energy costs, aggravated by wars, relatively high labor costs and the European Union’s trademark slow decision-making process are also blamed for blunting Europe’s competitive edge.

 

Amid the dire forecasts, Germany’s Federal Agency for Breakthrough Innovation — called SPRIND — offers a breath of optimism.

 

The office is determined to buck ingrained assumptions and prove that governments can fund ambitious, cutting-edge ventures quickly and across national borders.

 

SPRIND is hardly the first effort by the European Union or Britain aimed at jump-starting innovation and research. The European Union’s 95.5-billion-euro Horizon Europe program has been running since 2020, and a €175 billion budget had been proposed for the next phase, scheduled to commence in 2028.

 

Horizon’s track record has been mixed, with critics complaining about cumbersome applications, confusing rules and a preference for large corporations over disruptive enterprises.

 

SPRIND is much smaller, but it thinks big. Its goal is innovations that “solve a massive technological, social or environmental problem,” create “a completely new market” or change an existing one.

 

“SPRIND is completely different,” said Mr. Scrutton, who has applied for countless grants during nearly four decades of research at C3 Biotech and the University of Manchester, where he is a professor. Most have timelines that span several months or even years.

 

The average time between the filing of a grant application at SPRIND and the signing of a contract is 14 days, according to the agency.

 

A seminal report on the competitiveness crisis commissioned by the European Union and headed by Mario Draghi, a former prime minister of Italy, singled out SPRIND’s approach as a model for innovation.

 

One inspiration for SPRIND was the Defense Advanced Research Projects Agency in the United States, or DARPA, which laid the foundation for the internet and backed projects like the Saturn rocket, armed drones and biofeedback systems.

 

Public research funding often steers clear of projects that may have a big payoff but also a big possibility of failure and could take a long time to develop. Groundbreaking innovations frequently perish in what’s known as the “valley of death” because they can’t get money to finance the leap between early-stage research and commercial production.

 

Jano Costard, whose title at SPRIND is head of challenges, wants to give researchers the “freedom to really push the boundaries and do something that you wouldn’t get funding for probably through a conventional venture capital route.”

 

The strategy is to offer challenges to develop pioneering products and technologies, and then to invite teams of entrepreneurs, academics and researchers to compete. A jury of experts screens the proposals.

 

SPRIND also offers grants, investment and financing for innovative start-ups outside its own sponsored challenges.

 

Since it started in early 2020, SPRIND has received 4,740 submissions and funded 376 teams. To obtain the next round of funding, a team must demonstrate measurable progress. Failures are expected.

 

The goal of C3 Biotech’s three-year biomanufacturing challenge was to create a way to cheaply convert waste streams into products that could ultimately be produced on an industrial scale.

 

Other challenges funded by SPRIND involved the application of artificial intelligence to health data, the use of microgravity for drug development and more.

 

One unusual feature of how SPRIND works is that while its funding comes from the German government, the organization is willing to spend it all over Europe, including the Manchester, England, suburb where C3 Biotech is located.

 

“We need to create breakthroughs, and these might not just come from Germany,” Mr. Costard said. After all, if a significant advance could, say, reduce energy prices in Germany, then it’s money well spent.

 

Last month, SPRIND awarded grants worth up to €26.5 million to 10 teams from five countries to develop frontier artificial intelligence systems. At the end of 2028, the agency plans to help three teams that emerge as winners to raise €1 billion each from public and private investors.

 

Vinnova, a Swedish government agency that similarly seeds innovative start-ups, is teaming up with SPRIND on a new technology challenge: how to destroy small drones while safeguarding civilians from fallout.

 

Darja Isaksson, Vinnova’s director general, said that “we were really surprised” to learn that SPRIND was funding teams outside Germany. That was “one of the things that really caught my attention.”

 

Other countries have taken notice. Last month, France announced plans to create its own breakthrough innovation agency that would work with SPRIND. The Netherlands, too, is eyeing proposals to set up its own version of the German incubator.

 

At C3 Biotech, the team is tending to SPRIND’s investment. In the lab are plastic vats filled with hungry, salt-loving bacteria that are feeding on starch left over from the manufacture of potato chips.

 

The biological compound that this fermentation produces can be processed into acrylic glass and used in a mold or 3-D printer to manufacture thousands of products, from automobile headlamps to children’s toys.

 

A scientist cradled prized prototypes in his hand: a tiny black toy brick, a translucent blue heart and a yellow Lego-like figurine, all made from Citrasyn, the new chemical produced in their lab.

 

Establishing a cheap process to make chemicals from waste streams “has been the holy grail of biomanufacturing for many, many years,” Mr. Scrutton said. This month, he and his team celebrated a landmark: 180 days of continuous production.

 

C3 Biotech is now talking to corporations that are interested in commercializing the discovery.

 

Companies like C3 Biotech could provide the kind of innovations that undergird Europe’s future prosperity.

 

“That is exactly kind of the background against which SPRIND was created,” Mr. Costard said. “Our job now is to make sure that we are able to create those new industries.”” [1]

 

 

1. Europe Pushes for Moonshot Breakthroughs to Compete With China and U.S. Cohen, Patricia.  New York Times (Online) New York Times Company. Aug 19, 2026.