The Political Physics of PLA Drift

Xi Jinping’s purges of the People’s Liberation Army are acts of cultivation — pruning for control, not eliminating rivals. But the system he built may now operate beyond his design.

Across a decade of purges culminating in the 2025 reshuffle, Xi Jinping has treated the People's Liberation Army less as an institution than a system under calibration. His campaigns of "rectification" are not spasms of insecurity but acts of cultivation — pruning for control rather than elimination. What looks to outsiders like chaos follows an internal logic: a physics of drift and correction.

Each round follows the same choreography. A communiqué appears late in the evening. Names vanish from Party rolls. Portfolios are reassigned with clerical calm. The language of expulsion is antiseptic — "violations of discipline," "breaches of law" — yet beneath the euphemism lies a steady process of system maintenance. Over 13 years of repetition, the purge has become routine: not a rupture but the mechanism through which stability is produced.

In systems theory, feedback is the process by which an organization maintains equilibrium. When output diverges from input — when signals distort — the controller intervenes to restore alignment. Xi's method operates on the same principle. He sees corruption not as moral decay but signal drift: a condition in which institutions begin to serve their own interests rather than transmit the centre's intent faithfully. The remedy, in his logic, is mechanical — remove friction until only obedience remains.

That conception predates his presidency. In the early 1980s, as secretary to Defence Minister Geng Biao, Xi watched the military barter influence, posts, and factory shares as private assets. Geng, a revolutionary of the puritan school, tried to halt the practice by re-subordinating military enterprises to Party control but the effort collapsed amid the permissiveness of early reform. The habits of monetised loyalty endured, spreading through procurement offices and research institutes alike. For Xi, then a young aide, the lesson was enduring: corruption corroded command via drift — a loss of alignment between authority and execution. Rectification became, in his mind, the restoration of a clean channel between the centre and its instruments.

The Fourth Plenum of October 2025 displayed that logic at scale. Zhang Shengmin, the army's chief inquisitor, was elevated to vice-chair of the Central Military Commission, replacing the purged He Weidong. Eleven members of the Central Committee — roughly one in twenty — were replaced in a single session, and nine generals expelled, eight of them Committee members. It was the largest turnover since 2017 and roughly double the average rate under Hu Jintao. The rhythm of elimination now defines the tempo of governance: periodic correction as proof of vitality.

In the Rocket Force, the process has been more absolute. Zhang Fengzhong, head of its political-work department, became the tenth senior general placed under investigation in two years — effectively erasing the upper ranks. The command structure responsible for China's nuclear arsenal has been rebuilt almost from scratch. Officially, the campaign remains anti-corruption orientated but functionally it is a redesign of control pathways. Political commissars now dominate positions once held by operational officers, signalling that reliability outweighs competence as the organising principle.

A feedback system, however, must balance correction with responsiveness. Too little adjustment and deviation compounds; too much, and the controller amplifies its own noise. Xi's discipline campaigns illustrate the latter. His purges eliminate drift in the short term — obedience tightens, channels clear — but over-correction produces inertia. When the cost of initiative is uncertainty about the boundary of loyalty, the rational response is stasis. The system becomes efficient at negating motion but clumsy at generating it.

The rise of Zhang Shengmin epitomises this paradox. A lifelong commissar with limited operational background, Zhang's authority derives not from command experience but proximity to the core. His ascent tells every officer where real power resides and the more precisely subordinates attempt to anticipate Xi's expectations, the less initiative survives.

The purge mechanism compounds this through opacity. The criteria for removal are never stated; infractions are defined after the fact. In theory, this enforces universal discipline — no one is exempt. In practice, it creates information asymmetry: every officer knows obedience is necessary but cannot know if it is sufficient. Compliance becomes performative. Energy that once flowed into operational problem-solving is redirected into reading silence.

The result is a hierarchy of anticipation. Each node waits for the one above to signal safety before acting, producing latency throughout the chain of command. The organisation remains orderly yet inert like a perfectly balanced gyroscope — stable but unresponsive to external change. Xi's mechanical solution to drift therefore generates a subtler form of it: not defiance, but hesitation. The Pattern Beneath the Pattern

Yet the distribution of purges reveals something more than mechanical correction. When examined closely, the casualties cluster not randomly but along specific institutional lineages. He Weidong, helicoptered into the Central Military Commission just three years prior, became the first CMC vice chairman removed in over four decades. Admiral Miao Hua, promoted in 2017 to rejuvenate the command, followed. Both men shared a distinctive career trajectory: enlisted as ordinary soldiers at sixteen, rising through the ranks of frontline combat units without the benefit of elite pedigree. The pattern extends downward. General Lin Xiangyang, former Eastern Theater Commander, was expelled alongside He. So was General Wang Xiubin of the CMC Joint Operations Command Center. General Wang Houbing, installed as Rocket Force commander after the 2023 purge, lasted barely two years before arrest. These officers shared more than corruption charges—they shared institutional origins in units specialised in Taiwan operations, and they shared a common profile: meritocratic advancement from below rather than princeling inheritance from above. Meanwhile, Zhang Youxia, the CMC's first-ranked vice chairman, remains. His father was a Red Army hero of the 1920s and 1930s, equal in stature to Xi Jinping's own father. Zhang built his career on revolutionary pedigree and performance in the 1979 Vietnam border war. At the October 2025 Victory Day parade, he was the only uniformed military officer on the rostrum, seated in the second row directly behind the Politburo Standing Committee — a position typically reserved for retired Standing Committee members. The symbolism was unmistakable. The presiding officer at that parade was not one of the five theater commanders, as tradition dictates, but a relatively low-ranking two-star Air Force general. The officers leading each military component marched past without identification—a departure from previous parades where names and positions were announced. Three of the seven CMC seats remain vacant months after the purges. The new minister of defense has not been appointed to the CMC or State Council as tradition requires. Promotion processes appear frozen. This is not the clean output of a feedback loop. It is the signature of factional warfare conducted under the cover of anti-corruption discipline. The Weaponisation of System Design

Xi's anti-corruption architecture created the conditions for this outcome. By making vulnerability universal—by establishing that every officer has taken and given bribes, that anyone could be next—he destroyed the horizontal trust networks that traditionally buffered factional conflict. Guanxi, once a distributed system of reciprocal obligation, collapsed into vertical dependence on the core. Officers became isolated nodes, unable to coordinate, unable to verify which charges were substantive and which were instrumental. This was, perhaps, Xi's intent: to prevent coordination against him by ensuring no network could relax. But the system he built had a flaw. By rendering everyone vulnerable and destroying lateral communication, he created the perfect weapon not only for himself but for anyone positioned to wield it. The anti-corruption mechanism became a neutral technology—a tool that could be turned to purposes beyond its designer's control. The 2023 Rocket Force purge may have been the catalyst. The Rocket Force, traditionally led by princelings due to the political sensitivity of nuclear command, saw its entire senior leadership investigated and removed. He Weidong and his network of meritocratic officers appear to have used corruption charges—charges that were likely true, given endemic graft—to dismantle a princeling stronghold. Zhang Youxia, watching his faction's domain eliminated, may have recognised both threat and opportunity: if the anti-corruption mechanism could be used to purge princelings, it could equally be reversed against those who wielded it. When Xi failed to protect the Rocket Force princelings—whether unable or unwilling—Zhang appears to have drawn a conclusion about the limits of princeling immunity and the vulnerability of the meritocrats. The counter-offensive followed. By October 2025, the entire cohort of non-princeling generals who had risen through Taiwan-focused units was effectively neutralised. The targeting was too precise to be coincidental. He Weidong reportedly designed the military's operational response to Nancy Pelosi's 2022 Taiwan visit—an aggressive posture that American officials called an overreaction. His network represented not just a meritocratic challenge to princeling dominance but a distinct operational philosophy: younger, harder-line officers trained specifically for cross-strait conflict. Their removal was not random system maintenance but strategic pruning of a factional branch. Xi's role in this remains opaque. If he elevated He Weidong to advance a more aggressive Taiwan posture, why abandon him when the power struggle emerged? If Zhang convinced Xi that the upstarts threatened regime stability, why is there no trace of this logic in Xi's speeches or directives? If this was a policy dispute between hawks and caution, why did the chairman of the CMC not intervene to resolve it? Three possibilities present themselves. First, Xi has lost control of the purge mechanism he created, and Zhang now operates it independently. Second, Xi recognises the factional struggle but judges intervention too risky given Zhang's authority within the military. Third, Xi is complicit, having concluded that princeling dominance—however operationally suboptimal—is less threatening than meritocratic networks he cannot reliably monitor. Each interpretation reveals the same structural flaw. Xi built a system designed to prevent coordination against him by creating universal vulnerability and information asymmetry. But this architecture is indifferent to who deploys it. By destroying horizontal trust, Xi ensured no faction could coalesce against the centre—but he also ensured he could not see factional warfare when it occurred within the vertical column of command he preserved. The very opacity that protects him from collective challenge blinds him to individual manipulation of the system's levers. The system thus operates on two levels simultaneously. On the surface, it appears as Xi intended: periodic correction maintaining alignment through fear. But beneath, it enables precisely the kind of coordinated action it was meant to prevent—so long as that coordination flows through someone positioned close enough to the core to mask factional intent as regime maintenance. Zhang's proximity, his revolutionary credentials, his institutional weight all positioned him to do what scattered junior officers could not: use Xi's anti-corruption system for factional ends while Xi either cannot detect it or dares not acknowledge it.

The Cybernetic Failure In cybernetic terms, this represents a catastrophic failure of requisite variety—but not the one initially apparent. The problem is not simply that Xi's control system cannot process the diversity of operational signals that real command demands. It is that the control system itself has become multi-agent: Xi designed a single-controller feedback loop, but the mechanism now serves multiple controllers whose objectives may diverge. When Zhang purges He's network, is he correcting drift or creating it? From Xi's perspective—assuming Xi even perceives the distinction—both interpretations are plausible. The officers were corrupt; corruption is drift; removal restores alignment. The system gives no signal that factional elimination differs from routine discipline. Xi's mechanical solution to corruption has therefore become a mechanical blindness to power struggle. The two are indistinguishable within the system's own logic. This creates a doom loop more subtle than simple over-correction. Xi cannot stop purging without allowing networks to reform—but every purge may serve purposes beyond his own. He cannot distinguish between his corrections and Zhang's manoeuvres because both use identical language, identical procedures, identical justifications. The anti-corruption architecture produces perfect mimicry: factional warfare looks exactly like regime maintenance. The cost compounds across domains. A military command structure optimised for preventing coordination cannot distinguish between preventing coups and preventing competence. When meritocratic officers specialising in Taiwan operations are systematically removed, the system reads it as corruption elimination—but the operational effect is the same as if Xi had deliberately decided to sideline Taiwan hawks in favour of princeling caution. Strategic drift is now invisible to the control system because it is encoded as anti-drift. The Rocket Force Paradox The Rocket Force reveals this paradox at its sharpest. Since its creation in 2015, the branch has managed the dual task of nuclear deterrence and conventional strike. Its credibility depends on clarity — unbroken communication from political authority through command channels to launch units. By decapitating its senior leadership twice in two years, the system has ensured loyalty at the cost of continuity. But whose loyalty? The Rocket Force was traditionally led by princelings, sons of the Communist elite, precisely because nuclear command requires absolute regime security. When that command structure was purged in 2023, it was replaced by meritocratic officers from conventional forces. When those officers were purged in 2025, the pattern suggested not corruption elimination but factional restoration. The nuclear arsenal's command structure has been rebuilt twice—not to improve performance but to ensure the right lineage controls it. A nuclear deterrent tolerates little ambiguity. Promotions delayed by investigation create transient gaps in authority. Testing schedules slip as newly promoted officers await confirmation from superiors themselves under review. In a crisis, hesitation can be misread as intent. But the system cannot distinguish between hesitation caused by procedural paralysis and hesitation caused by factional recalibration. Both produce silence. Both degrade deterrence. The cybernetic controller—Xi—receives no signal differentiating the two. The strategic cost is identical either way, but the political logic diverges completely. If the Rocket Force paralysis results from excessive anti-corruption zeal, Xi might moderate the campaign. If it results from factional warfare he cannot or will not resolve, moderation is impossible. The system is trapped by its own design: it produces the problem, it prevents detection of the problem, and it prevents solution of the problem. The Stalin Precedent Revisited The parallel to Stalin's 1937 purge of the Red Army general staff now appears more precise. Stalin secured political loyalty but annihilated initiative—a trade-off he judged acceptable because a competent military posed greater regime risk than external threat. Xi's version appeared less violent but structurally similar: a high-frequency correction loop sacrificing adaptability for alignment. But the deeper parallel may be this: Stalin's purge was not entirely his own. It was enabled and amplified by subordinates—Yezhov, Beria—who used the mechanism for their own factional ends while Stalin either could not detect it or chose not to stop it. The purge acquired momentum beyond Stalin's initial intent. Officers denounced rivals. Interrogators extracted confessions implicating others. The system fed on itself. Xi's purge mechanism shows the same autocatalytic property. Once corruption charges become the universal solvent for removing rivals, every ambitious officer has incentive to deploy them. Zhang may have initiated the pattern, but others will learn it. The system, designed to centralise control, becomes a distributed weapon. Xi wanted one controller and many subjects; he has created multiple controllers masked as enforcers. Fear replaces friction as the system's organising medium—but fear of what? Not simply fear of Xi. Fear of whoever can credibly deploy the purge mechanism. In an environment of perfect opacity, officers cannot distinguish between Xi's wrath and Zhang's manoeuvring. Both manifest identically: a late-night communiqué, a name vanishing from Party rolls, antiseptic language. The rational response is to anticipate both—to seek protection not just from Xi but from anyone who might turn the weapon against them. For decades, the PLA's cohesion rested on reciprocal networks that distributed trust horizontally when formal institutions faltered. These networks were inefficient but predictable: most officers knew who was connected to whom. Xi compressed that field into a single vertical channel, assuming he would control it. But verticality is not singularity. Zhang has demonstrated that vertical channels can carry multiple signals—that the same conduit meant to transmit Xi's discipline can carry factional warfare masked as discipline. The formal architecture remains intact, but its informal circuits—the redundancies that make large organisations resilient—have been disabled. Except one redundancy was not eliminated but hidden: the princeling network, which predates Xi, which shares revolutionary legitimacy with Xi, which cannot be purged without destabilising the entire basis of Party authority. Zhang represents not a rebel faction but a structural feature—an aristocracy the system was designed to preserve even as it claimed to eliminate all networks. Drift as Equilibrium Drift, then, is not moral decay but systemic lag—the distance between command intent and functional response. It can arise from corruption, when incentives diverge from purpose. It can arise from over-control, when initiative collapses under surveillance. Or it can arise from multi-agent control, when the mechanism meant to prevent drift becomes the vector for creating it. Xi's project to eliminate the first form of drift has produced the second and enabled the third. The PLA behaves less like a disciplined hierarchy than a contested system: multiple actors deploying the same control mechanism for divergent ends, none able to coordinate openly, all operating in a fog of indistinguishable signals. The result is not chaos but a peculiar form of order—movement in place, energy that recirculates without producing motion. The paradox resolves when understood not as Xi's failure but the system's success at serving multiple masters. Zhang has not broken the system; he has discovered how to use it. The perpetual purge continues not because Xi designed it that way but because it now serves functions beyond his design. Each dismissal reaffirms conditional tenure, dissolves emerging networks, and reminds subordinates that security is provisional. But it also allows factional warfare to continue indefinitely under the cover of discipline. If this interpretation holds, Xi faces an impossible choice. He cannot stop purging without allowing networks to reform—including the meritocratic networks he has worked to dismantle. He cannot continue purging without enabling Zhang and others to use the mechanism for their own ends. He cannot reform the system to distinguish between his corrections and others' manipulations without acknowledging that the system operates beyond his control—an admission that would itself threaten his authority. The system is thus trapped in a stable dysfunctional equilibrium. It produces the appearance of control while enabling the reality of factional struggle. It prevents collective coordination against Xi while allowing individual actors near the core to coordinate their own agendas. It maintains Xi's dominance while hollowing out his capacity to direct outcomes. The perpetual purge becomes not evidence of strength or weakness but of a system that has achieved a form of homeostasis—a steady state of controlled instability that serves multiple functions for multiple actors, none of whom can afford to stop it. The Strategic Residue Externally, the consequences remain unpredictable but the pattern clarifies. The systematic removal of officers specialised in Taiwan operations—regardless of whether this was Xi's intent or Zhang's achievement—has strategic implications independent of factional logic. The PLA's Taiwan-focused command networks have been decapitated twice in two years. The officers who designed aggressive responses to American provocations are gone. The institutional memory of cross-strait operational planning now resides in mid-tier officers who have watched their superiors vanish for reasons never specified. A military operating under these conditions may respond to crises not with aggression but with delay—a form of uncertainty foreign analysts often misread as calculation. In truth, hesitation may reflect not prudence but procedural paralysis: no one willing to act without explicit command, and explicit command delayed while the command structure itself remains contested. Zhang's ascendancy may have secured princeling control, but it has not restored operational clarity. The system remains frozen, awaiting signals that are themselves ambiguous.

Deterrence depends on credible communication of intent. This system produces silence—not the silence of strength but the silence of internal incoherence. When mid-level officers cannot determine whether caution or aggression better serves survival, when senior commanders dare not advocate policies that might be read as factional positioning, when the nuclear command structure has been rebuilt twice in two years for reasons never publicly explained, the result is not a disciplined instrument but an opaque one. The opacity that protects Xi from internal challenge can, in conflict, obscure his own intentions from those tasked to execute them. If Zhang's factional victory has shifted the balance toward caution, Xi may find his directives implemented through a filter he did not authorise. If meritocratic networks in lower ranks persist despite purges above, he may face the opposite problem: aggressive implementation of ambiguous guidance. The system gives him no reliable signal of which dynamic prevails. The Machine That Cannot Stop Ultimately, Xi's politics reduce governance to mechanics—power sustained by subtraction, pruning until nothing moves without permission. But the machine he built to prevent motion has acquired its own momentum. It cannot distinguish between his commands and others' manipulations. It cannot stop without risking the networks it was meant to prevent. It cannot continue without deepening the incapacity that control was meant to prevent. The system survives by remaining forever on the edge of dysfunction, suspended between motion and paralysis. But edge states are unstable in ways their designers cannot predict. By creating universal vulnerability, Xi enabled factional warfare. By destroying horizontal trust, he blinded himself to its operation. By insisting on mechanical discipline, he produced a mechanism that serves multiple purposes—some his, some not, all indistinguishable. The question is not whether this system can endure. Systems optimised for survival often do, regardless of performance. The question is what it sacrifices to endure and who makes those trade-offs. A military incapable of initiative may serve Xi's political needs perfectly—or it may serve Zhang's vision of cautious princeling dominance. The opacity makes both possible, neither verifiable. Xi has built a machine that can neither stop nor accelerate, that prevents coordination while enabling manipulation, that centralises authority while distributing the means of factional warfare. Whether this was his intent or his failure may be irrelevant. The system now operates according to its own logic—a logic of perpetual correction that feeds on the drift it endlessly reproduces, serving masters who may not even know they compete for control of the same mechanism. The purge continues not because Xi wills it but because the system requires it. And in that requirement lies the deepest irony: the control mechanism designed to ensure perfect obedience has produced a machine that no one fully controls—including its architect.